update server

model/model_config.py

@@ -99,6 +99,12 @@ register_model_config(
     online_model=False
 )
 
+register_model_config(
+    model_name="magic123",
+    i2s_model=True,
+    online_model=False
+)
+
 register_model_config(
     model_name="openlrm",
     i2s_model=True,

model/model_worker.py

@@ -45,9 +45,14 @@ class BaseModelWorker:
         galley = "image2shape" if self.i2s_model else "text2shape"
         rgb_name = f"{galley}_{self.model_name}_{offline_idx}_rgb.gif"
         normal_name = f"{galley}_{self.model_name}_{offline_idx}_normal.gif"
+        geo_name = f"{galley}_{self.model_name}_{offline_idx}_geo.gif"
+
         rgb_url = f"{GIF_SERVER}/{rgb_name}"
         normal_url = f"{GIF_SERVER}/{normal_name}"
-        return {'rgb': rgb_url, 'normal': normal_url}
+        geo_url = f"{GIF_SERVER}/{geo_name}"
+        return {'rgb': rgb_url, 
+                'normal': normal_url,
+                'geo': geo_url}
 
     def inference(self, prompt):
         pass

serve/gradio_web_i2s.py

@@ -32,12 +32,12 @@ def build_i2s_ui_side_by_side_anony(models):
 # ⚔️  3DGen-Arena ⚔️ : Benchmarking Image-to-3D generative models 
 ## 📜 Rules
 - Upload image to two anonymous models in same area and vote for the better one!
-- When the results are ready, click the button below to vote.
-- Vote won't be counted if model identity is revealed during conversation.
+- When the results are ready, click the buttons below to vote.
+- Until all dimensions have been voted, the anonymous models are displayed.
 - Click "Clear" to start a new round.
 
 ## 🏆 Arena Elo 
-Find out who is the 🥇conditional image generation models! More models are going to be supported.
+Find out who is the 🥇 image-to-3D generation models! More models are going to be supported.
 
 ## 👇 Generating now!
 
@@ -56,24 +56,51 @@ Find out who is the 🥇conditional image generation models! More models are goi
             model_description_md = get_model_description_md(model_list)
             gr.Markdown(model_description_md, elem_id="model_description_markdown")
 
+        with gr.Column():
+            with gr.Row():
+                geo_left = gr.Image(label = "Geometry A", 
+                    interactive=False, show_download_button=True)
+                normal_left = gr.Image(label = "Normal A", 
+                    interactive=False, show_download_button=True)
+                rgb_left = gr.Image(label = "RGB A", 
+                    interactive=False, show_download_button=True)
+            with gr.Row():
+                geo_right = gr.Image(label = "Geometry B", 
+                    interactive=False, show_download_button=True)
+                normal_right = gr.Image(label = "Normal B",
+                    interactive=False, show_download_button=True,)
+                rgb_right = gr.Image(label = "RGB B",
+                    interactive=False, show_download_button=True,)
+
+        with gr.Row():
+            # with gr.Column(scale=0.8):
+            #     gr.Markdown("", visible=False)
+            with gr.Column():
+                model_selector_left =gr.Markdown("", visible=False)
+            with gr.Column():
+                model_selector_right = gr.Markdown("", visible=False)
+        with gr.Row():
+            slow_warning = gr.Markdown("", elem_id="notice_markdown") 
+
         with gr.Row():
-            with gr.Column(scale=0.8):
+            with gr.Column():
                 imagebox = gr.Image(
-                    width=512,
                     show_label=False,
                     visible=True,
                     interactive=False,   ## only support offline
                     label = "Prompt",
                     elem_id="input_box",
-                )        
-                
+                )
+                sample_btn = gr.Button(value="🎲 Sample", variant="primary")
+            
+            with gr.Column(scale=3.0):
                 plausive_md = gr.Markdown("Geometry Plausibility: ", visible=False, elem_id="evaldim_markdown")
                 with gr.Row(elem_id="Geometry Plausibility"):
                     plausive_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
+                        value="👆  A is better", visible=False, interactive=False
                     )
                     plausive_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
+                        value="👇  B is better", visible=False, interactive=False
                     )
                     plausive_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
                     plausive_bothbad_btn = gr.Button(
@@ -83,10 +110,10 @@ Find out who is the 🥇conditional image generation models! More models are goi
                 geo_md = gr.Markdown("Geometry Details: ", visible=False, elem_id="evaldim_markdown")
                 with gr.Row(elem_id="Geometry Details"):
                     geo_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
+                        value="👆  A is better", visible=False, interactive=False
                     )
                     geo_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
+                        value="👇  B is better", visible=False, interactive=False
                     )
                     geo_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
                     geo_bothbad_btn = gr.Button(
@@ -96,10 +123,10 @@ Find out who is the 🥇conditional image generation models! More models are goi
                 text_md = gr.Markdown("Texture Quality: ", visible=False, elem_id="evaldim_markdown")
                 with gr.Row(elem_id="Texture Quality"):
                     text_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
+                        value="👆  A is better", visible=False, interactive=False
                     )
                     text_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
+                        value="👇  B is better", visible=False, interactive=False
                     )
                     text_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
                     text_bothbad_btn = gr.Button(
@@ -109,10 +136,10 @@ Find out who is the 🥇conditional image generation models! More models are goi
                 coherence_md = gr.Markdown("Geometry-Texture Coherency: ", visible=False, elem_id="evaldim_markdown")
                 with gr.Row(elem_id="Geometry-Texture Coherency"):
                     coherence_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
+                        value="👆  A is better", visible=False, interactive=False
                     )
                     coherence_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
+                        value="👇  B is better", visible=False, interactive=False
                     )
                     coherence_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
                     coherence_bothbad_btn = gr.Button(
@@ -122,108 +149,16 @@ Find out who is the 🥇conditional image generation models! More models are goi
                 align_md = gr.Markdown("Visual Alignment: ", visible=False, elem_id="evaldim_markdown")
                 with gr.Row(elem_id="Visual Alignment"):
                     align_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
+                        value="👆  A is better", visible=False, interactive=False
                     )
                     align_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
+                        value="👇  B is better", visible=False, interactive=False
                     )
                     align_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
                     align_bothbad_btn = gr.Button(
                         value="👎  Both are bad", visible=False, interactive=False
                     )
 
-
-                sample_btn = gr.Button(value="🎲 Sample", variant="primary")
-                # send_btn = gr.Button(value="📤 Send", variant="primary")
-                # clear_btn = gr.Button(value="🗑️ Clear", interactive=False)
-                # regenerate_btn = gr.Button(value="🔄  Regenerate", interactive=False)
-                # share_btn = gr.Button(value="📷  Share")    
-            with gr.Column():
-                normal_left = gr.Image(label = "Normal A", 
-                    interactive=False, show_download_button=True)
-                rgb_left = gr.Image(label = "RGB A", 
-                    interactive=False, show_download_button=True)
-            with gr.Column():
-                normal_right = gr.Image(label = "Normal B",
-                    interactive=False, show_download_button=True,)
-                rgb_right = gr.Image(label = "RGB B",
-                    interactive=False, show_download_button=True,)
-
-        with gr.Row():
-            with gr.Column(scale=0.8):
-                gr.Markdown("", visible=False)
-            with gr.Column():
-                model_selector_left =gr.Markdown("", visible=False)
-            with gr.Column():
-                model_selector_right = gr.Markdown("", visible=False)
-        with gr.Row():
-            slow_warning = gr.Markdown("", elem_id="notice_markdown") 
-
-        # with gr.Row(elem_id="Geometry Plausibility"):
-        #     plausive_md = gr.Markdown("Geometry Plausibility: ", visible=False, elem_id="evaldim_markdown")
-        #     plausive_leftvote_btn = gr.Button(
-        #         value="👈  A is better", visible=False, interactive=False
-        #     )
-        #     plausive_rightvote_btn = gr.Button(
-        #         value="👉  B is better", visible=False, interactive=False
-        #     )
-        #     plausive_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-        #     plausive_bothbad_btn = gr.Button(
-        #         value="👎  Both are bad", visible=False, interactive=False
-        #     )
-
-        # with gr.Row(elem_id="Geometry Details"):
-        #     geo_md = gr.Markdown("Geometry Details: ", visible=False, elem_id="evaldim_markdown")
-        #     geo_leftvote_btn = gr.Button(
-        #         value="👈  A is better", visible=False, interactive=False
-        #     )
-        #     geo_rightvote_btn = gr.Button(
-        #         value="👉  B is better", visible=False, interactive=False
-        #     )
-        #     geo_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-        #     geo_bothbad_btn = gr.Button(
-        #         value="👎  Both are bad", visible=False, interactive=False
-        #     )
-
-        # with gr.Row(elem_id="Texture Quality"):
-        #     text_md = gr.Markdown("Texture Quality: ", visible=False, elem_id="evaldim_markdown")
-        #     text_leftvote_btn = gr.Button(
-        #         value="👈  A is better", visible=False, interactive=False
-        #     )
-        #     text_rightvote_btn = gr.Button(
-        #         value="👉  B is better", visible=False, interactive=False
-        #     )
-        #     text_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-        #     text_bothbad_btn = gr.Button(
-        #         value="👎  Both are bad", visible=False, interactive=False
-        #     )
-        
-        # with gr.Row(elem_id="Geometry-Texture Coherency"):
-        #     coherence_md = gr.Markdown("Geometry-Texture Coherency: ", visible=False, elem_id="evaldim_markdown")
-        #     coherence_leftvote_btn = gr.Button(
-        #         value="👈  A is better", visible=False, interactive=False
-        #     )
-        #     coherence_rightvote_btn = gr.Button(
-        #         value="👉  B is better", visible=False, interactive=False
-        #     )
-        #     coherence_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-        #     coherence_bothbad_btn = gr.Button(
-        #         value="👎  Both are bad", visible=False, interactive=False
-        #     )
-
-        # with gr.Row(elem_id="Visual Alignment"):
-        #     align_md = gr.Markdown("Visual Alignment: ", visible=False, elem_id="evaldim_markdown")
-        #     align_leftvote_btn = gr.Button(
-        #         value="👈  A is better", visible=False, interactive=False
-        #     )
-        #     align_rightvote_btn = gr.Button(
-        #         value="👉  B is better", visible=False, interactive=False
-        #     )
-        #     align_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-        #     align_bothbad_btn = gr.Button(
-        #         value="👎  Both are bad", visible=False, interactive=False
-        #     )
-
     with gr.Row():
         # sample_btn = gr.Button(value="🎲 Sample", variant="primary")
         clear_btn = gr.Button(value="🗑️ Clear", interactive=False)
@@ -235,7 +170,7 @@ Find out who is the 🥇conditional image generation models! More models are goi
 
     states = [state_0, state_1]
     model_selectors = [model_selector_left, model_selector_right]
-    results = [normal_left, rgb_left, normal_right, rgb_right]
+    results = [geo_left, normal_left, rgb_left, geo_right, normal_right, rgb_right]
 
     geo_btn_list = [geo_leftvote_btn, geo_rightvote_btn, geo_tie_btn, geo_bothbad_btn]
     text_btn_list = [text_leftvote_btn, text_rightvote_btn, text_tie_btn, text_bothbad_btn]
@@ -389,8 +324,8 @@ Find out who is the 🥇conditional image generation models! More models are goi
             gr.Markdown(model_description_md, elem_id="model_description_markdown")
 
         with gr.Row():
-            with gr.Column(scale=0.8):
-                gr.Markdown("Select Players:", elem_id="evaldim_markdown")
+            # with gr.Column(scale=0.8):
+            #     gr.Markdown("Select Players:", elem_id="evaldim_markdown")
             with gr.Column():
                 model_selector_left = gr.Dropdown(
                     choices=model_list,
@@ -408,23 +343,35 @@ Find out who is the 🥇conditional image generation models! More models are goi
                     container=False,
                     )
 
+        with gr.Column():
+            with gr.Row():
+                geo_left = gr.Image(label = "Geometry A", interactive=False, show_download_button=True)
+                normal_left = gr.Image(label = "Normal A", interactive=False, show_download_button=True)
+                rgb_left = gr.Image(label = "RGB A", interactive=False, show_download_button=True)
+            with gr.Row():
+                geo_right = gr.Image(label = "Geometry B", interactive=False, show_download_button=True)
+                normal_right = gr.Image(label = "Normal B", interactive=False, show_download_button=True,)
+                rgb_right = gr.Image(label = "RGB B", interactive=False, show_download_button=True,)
+
         with gr.Row():
-            with gr.Column(scale=0.8):
+            with gr.Column():
                 imagebox = gr.Image(
-                    width=512,
                     show_label=False,
                     visible=True,
                     interactive=False,   ## only support offline
+                    label = "Prompt",
                     elem_id="input_box",
                 )
-
+                sample_btn = gr.Button(value="🎲 Sample", variant="primary")
+            
+            with gr.Column(scale=3.0):
                 plausive_md = gr.Markdown("Geometry Plausibility: ", visible=False, elem_id="evaldim_markdown")
                 with gr.Row(elem_id="Geometry Plausibility"):
                     plausive_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
+                        value="👆  A is better", visible=False, interactive=False
                     )
                     plausive_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
+                        value="👇  B is better", visible=False, interactive=False
                     )
                     plausive_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
                     plausive_bothbad_btn = gr.Button(
@@ -434,10 +381,10 @@ Find out who is the 🥇conditional image generation models! More models are goi
                 geo_md = gr.Markdown("Geometry Details: ", visible=False, elem_id="evaldim_markdown")
                 with gr.Row(elem_id="Geometry Details"):
                     geo_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
+                        value="👆  A is better", visible=False, interactive=False
                     )
                     geo_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
+                        value="👇  B is better", visible=False, interactive=False
                     )
                     geo_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
                     geo_bothbad_btn = gr.Button(
@@ -447,10 +394,10 @@ Find out who is the 🥇conditional image generation models! More models are goi
                 text_md = gr.Markdown("Texture Quality: ", visible=False, elem_id="evaldim_markdown")
                 with gr.Row(elem_id="Texture Quality"):
                     text_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
+                        value="👆  A is better", visible=False, interactive=False
                     )
                     text_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
+                        value="👇  B is better", visible=False, interactive=False
                     )
                     text_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
                     text_bothbad_btn = gr.Button(
@@ -460,10 +407,10 @@ Find out who is the 🥇conditional image generation models! More models are goi
                 coherence_md = gr.Markdown("Geometry-Texture Coherency: ", visible=False, elem_id="evaldim_markdown")
                 with gr.Row(elem_id="Geometry-Texture Coherency"):
                     coherence_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
+                        value="👆  A is better", visible=False, interactive=False
                     )
                     coherence_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
+                        value="👇  B is better", visible=False, interactive=False
                     )
                     coherence_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
                     coherence_bothbad_btn = gr.Button(
@@ -473,32 +420,23 @@ Find out who is the 🥇conditional image generation models! More models are goi
                 align_md = gr.Markdown("Visual Alignment: ", visible=False, elem_id="evaldim_markdown")
                 with gr.Row(elem_id="Visual Alignment"):
                     align_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
+                        value="👆  A is better", visible=False, interactive=False
                     )
                     align_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
+                        value="👇  B is better", visible=False, interactive=False
                     )
                     align_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
                     align_bothbad_btn = gr.Button(
                         value="👎  Both are bad", visible=False, interactive=False
                     )
 
-                sample_btn = gr.Button(value="🎲 Sample", variant="primary")
-
-            with gr.Column():
-                normal_left = gr.Image(width=512, label = "Normal A", show_download_button=True)
-                rgb_left = gr.Image(width=512, label = "RGB A", show_download_button=True)
-            with gr.Column():
-                normal_right = gr.Image(width=512, label = "Normal B", show_download_button=True,)
-                rgb_right = gr.Image(width=512, label = "RGB B", show_download_button=True,)
-
         # with gr.Row(elem_id="Geometry Plausibility"):
         #     plausive_md = gr.Markdown("Geometry Plausibility: ", visible=False, elem_id="evaldim_markdown")
         #     plausive_leftvote_btn = gr.Button(
-        #         value="👈  A is better", visible=False, interactive=False
+        #         value="👆  A is better", visible=False, interactive=False
         #     )
         #     plausive_rightvote_btn = gr.Button(
-        #         value="👉  B is better", visible=False, interactive=False
+        #         value="👇  B is better", visible=False, interactive=False
         #     )
         #     plausive_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
         #     plausive_bothbad_btn = gr.Button(
@@ -508,10 +446,10 @@ Find out who is the 🥇conditional image generation models! More models are goi
         # with gr.Row(elem_id="Geometry Details"):
         #     geo_md = gr.Markdown("Geometry Details: ", visible=False, elem_id="evaldim_markdown")
         #     geo_leftvote_btn = gr.Button(
-        #         value="👈  A is better", visible=False, interactive=False
+        #         value="👆  A is better", visible=False, interactive=False
         #     )
         #     geo_rightvote_btn = gr.Button(
-        #         value="👉  B is better", visible=False, interactive=False
+        #         value="👇  B is better", visible=False, interactive=False
         #     )
         #     geo_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
         #     geo_bothbad_btn = gr.Button(
@@ -521,10 +459,10 @@ Find out who is the 🥇conditional image generation models! More models are goi
         # with gr.Row(elem_id="Texture Quality"):
         #     text_md = gr.Markdown("Texture Quality: ", visible=False, elem_id="evaldim_markdown")
         #     text_leftvote_btn = gr.Button(
-        #         value="👈  A is better", visible=False, interactive=False
+        #         value="👆  A is better", visible=False, interactive=False
         #     )
         #     text_rightvote_btn = gr.Button(
-        #         value="👉  B is better", visible=False, interactive=False
+        #         value="👇  B is better", visible=False, interactive=False
         #     )
         #     text_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
         #     text_bothbad_btn = gr.Button(
@@ -534,10 +472,10 @@ Find out who is the 🥇conditional image generation models! More models are goi
         # with gr.Row(elem_id="Geometry-Texture Coherency"):
         #     coherence_md = gr.Markdown("Geometry-Texture Coherency: ", visible=False, elem_id="evaldim_markdown")
         #     coherence_leftvote_btn = gr.Button(
-        #         value="👈  A is better", visible=False, interactive=False
+        #         value="👆  A is better", visible=False, interactive=False
         #     )
         #     coherence_rightvote_btn = gr.Button(
-        #         value="👉  B is better", visible=False, interactive=False
+        #         value="👇  B is better", visible=False, interactive=False
         #     )
         #     coherence_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
         #     coherence_bothbad_btn = gr.Button(
@@ -547,10 +485,10 @@ Find out who is the 🥇conditional image generation models! More models are goi
         # with gr.Row(elem_id="Visual Alignment"):
         #     align_md = gr.Markdown("Visual Alignment: ", visible=False, elem_id="evaldim_markdown")
         #     align_leftvote_btn = gr.Button(
-        #         value="👈  A is better", visible=False, interactive=False
+        #         value="👆  A is better", visible=False, interactive=False
         #     )
         #     align_rightvote_btn = gr.Button(
-        #         value="👉  B is better", visible=False, interactive=False
+        #         value="👇  B is better", visible=False, interactive=False
         #     )
         #     align_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
         #     align_bothbad_btn = gr.Button(
@@ -567,7 +505,7 @@ Find out who is the 🥇conditional image generation models! More models are goi
 
     states = [state_0, state_1]
     model_selectors = [model_selector_left, model_selector_right]
-    results = [normal_left, rgb_left, normal_right, rgb_right]
+    results = [geo_left, normal_left, rgb_left, geo_right, normal_right, rgb_right]
 
     geo_btn_list = [geo_leftvote_btn, geo_rightvote_btn, geo_tie_btn, geo_bothbad_btn]
     text_btn_list = [text_leftvote_btn, text_rightvote_btn, text_tie_btn, text_bothbad_btn]
@@ -747,48 +685,52 @@ def build_i2s_ui_single_model(models):
             show_label=False
         )
 
+    with gr.Row():
+        geo = gr.Image(label="Geometry", interactive=False, show_download_button=True)
+        normal = gr.Image(label="Normal", interactive=False, show_download_button=True)
+        rgb = gr.Image(label="RGB", interactive=False, show_download_button=True)
+
     with gr.Row():
         with gr.Column():
             imagebox = gr.Image(
                 show_label=False,
                 visible=True,
                 interactive=False,   ## only support offline
+                label = "Prompt",
                 elem_id="input_box",
-                label="Prompt",
             )
             sample_btn = gr.Button(value="🎲 Sample", variant="primary")
-        normal = gr.Image(width=512, label = "Normal", show_download_button=True)
-        rgb = gr.Image(width=512, label = "RGB", show_download_button=True,)
-
-    with gr.Row(elem_id="Geometry Plausibility"):
-        plausive_md = gr.Markdown("Geometry Plausibility: ", elem_id="evaldim_markdown")
-        plausive_upvote_btn = gr.Button(value="👍  Upvote", interactive=False)
-        plausive_downvote_btn = gr.Button(value="👎  Downvote", interactive=False)
-        plausive_flag_btn = gr.Button(value="⚠️  Flag", interactive=False)
         
-    with gr.Row(elem_id="Geometry Details"):
-        geo_md = gr.Markdown("Geometry Details: ", elem_id="evaldim_markdown")
-        geo_upvote_btn = gr.Button(value="👍  Upvote", interactive=False)
-        geo_downvote_btn = gr.Button(value="👎  Downvote", interactive=False)
-        geo_flag_btn = gr.Button(value="⚠️  Flag", interactive=False)
-
-    with gr.Row(elem_id="Texture Quality"):
-        text_md = gr.Markdown("Texture Quality: ", elem_id="evaldim_markdown")
-        text_upvote_btn = gr.Button(value="👍  Upvote", interactive=False)
-        text_downvote_btn = gr.Button(value="👎  Downvote", interactive=False)
-        text_flag_btn = gr.Button(value="⚠️  Flag", interactive=False)
-    
-    with gr.Row(elem_id="Geometry-Texture Coherency"):
-        coherence_md = gr.Markdown("Geometry-Texture Coherency: ", elem_id="evaldim_markdown")
-        coherence_upvote_btn = gr.Button(value="👍  Upvote", interactive=False)
-        coherence_downvote_btn = gr.Button(value="👎  Downvote", interactive=False)
-        coherence_flag_btn = gr.Button(value="⚠️  Flag", interactive=False)
-
-    with gr.Row(elem_id="Visual Alignment"):
-        align_md = gr.Markdown("Visual Alignment: ", elem_id="evaldim_markdown")
-        align_upvote_btn = gr.Button(value="👍  Upvote", interactive=False)
-        align_downvote_btn = gr.Button(value="👎  Downvote", interactive=False)
-        align_flag_btn = gr.Button(value="⚠️  Flag", interactive=False)
+        with gr.Column(scale=2.0):
+            plausive_md = gr.Markdown("Geometry Plausibility: ", elem_id="evaldim_markdown")
+            with gr.Row(elem_id="Geometry Plausibility"):
+                plausive_upvote_btn = gr.Button(value="👍  Upvote", interactive=False)
+                plausive_downvote_btn = gr.Button(value="👎  Downvote", interactive=False)
+                plausive_flag_btn = gr.Button(value="⚠️  Flag", interactive=False)
+                
+            geo_md = gr.Markdown("Geometry Details: ", elem_id="evaldim_markdown")
+            with gr.Row(elem_id="Geometry Details"):
+                geo_upvote_btn = gr.Button(value="👍  Upvote", interactive=False)
+                geo_downvote_btn = gr.Button(value="👎  Downvote", interactive=False)
+                geo_flag_btn = gr.Button(value="⚠️  Flag", interactive=False)
+
+            text_md = gr.Markdown("Texture Quality: ", elem_id="evaldim_markdown")
+            with gr.Row(elem_id="Texture Quality"):
+                text_upvote_btn = gr.Button(value="👍  Upvote", interactive=False)
+                text_downvote_btn = gr.Button(value="👎  Downvote", interactive=False)
+                text_flag_btn = gr.Button(value="⚠️  Flag", interactive=False)
+            
+            coherence_md = gr.Markdown("Geometry-Texture Coherency: ", elem_id="evaldim_markdown")
+            with gr.Row(elem_id="Geometry-Texture Coherency"):
+                coherence_upvote_btn = gr.Button(value="👍  Upvote", interactive=False)
+                coherence_downvote_btn = gr.Button(value="👎  Downvote", interactive=False)
+                coherence_flag_btn = gr.Button(value="⚠️  Flag", interactive=False)
+
+            align_md = gr.Markdown("Visual Alignment: ", elem_id="evaldim_markdown")
+            with gr.Row(elem_id="Visual Alignment"):
+                align_upvote_btn = gr.Button(value="👍  Upvote", interactive=False)
+                align_downvote_btn = gr.Button(value="👎  Downvote", interactive=False)
+                align_flag_btn = gr.Button(value="⚠️  Flag", interactive=False)
 
 
     with gr.Row():
@@ -830,7 +772,7 @@ def build_i2s_ui_single_model(models):
     model_selector.change(
         clear_i2s_history, 
         None, 
-        [state, imagebox, normal, rgb], 
+        [state, imagebox, geo, normal, rgb], 
         api_name="clear_history_single",
         show_progress="full"
     ).then(
@@ -847,17 +789,17 @@ def build_i2s_ui_single_model(models):
     ).then(
         reset_state,
         state,
-        [state, normal, rgb] + vote_btn_list + [regenerate_btn, clear_btn]
+        [state, geo, normal, rgb] + vote_btn_list + [regenerate_btn, clear_btn]
     )
 
     imagebox.upload(
         reset_state,
         state,
-        [state, normal, rgb] + vote_btn_list + [regenerate_btn, clear_btn]
+        [state, geo, normal, rgb] + vote_btn_list + [regenerate_btn, clear_btn]
     ).then(
         gen_func,
         [state, imagebox, model_selector],
-        [state, normal, rgb],
+        [state, geo, normal, rgb],
         api_name="submit_btn_single",
         show_progress = "full"
     ).then(
@@ -869,11 +811,11 @@ def build_i2s_ui_single_model(models):
     send_btn.click(
         reset_state,
         state,
-        [state, normal, rgb] + vote_btn_list + [regenerate_btn, clear_btn]
+        [state, geo, normal, rgb] + vote_btn_list + [regenerate_btn, clear_btn]
     ).then(
         gen_func,
         [state, imagebox, model_selector],
-        [state, normal, rgb],
+        [state, geo, normal, rgb],
         api_name="send_btn_single",
         show_progress = "full"
     ).then(
@@ -885,7 +827,7 @@ def build_i2s_ui_single_model(models):
     clear_btn.click(
         clear_i2s_history, 
         None, 
-        [state, imagebox, normal, rgb], 
+        [state, imagebox, geo, normal, rgb], 
         api_name="clear_history_single",
         show_progress="full"
     ).then(
@@ -897,11 +839,11 @@ def build_i2s_ui_single_model(models):
     regenerate_btn.click(
         reset_state,
         state,
-        [state, normal, rgb] + vote_btn_list + [regenerate_btn, clear_btn]
+        [state, geo, normal, rgb] + vote_btn_list + [regenerate_btn, clear_btn]
     ).then(
         gen_func,
         [state, imagebox, model_selector],
-        [state, normal, rgb],
+        [state, geo, normal, rgb],
         api_name="regenerate_btn_single",
         show_progress = "full"
     ).then(

serve/gradio_web_t2s.py

@@ -56,94 +56,19 @@ Find out who is the 🥇conditional image generation models! More models are goi
         with gr.Accordion("🔍 Expand to see all Arena players", open=False):
             model_description_md = get_model_description_md(model_list)
             gr.Markdown(model_description_md, elem_id="model_description_markdown")
-        with gr.Row():
-            with gr.Column():
-                textbox = gr.Textbox(
-                    show_label=False,
-                    placeholder="👉 Enter your prompt or Sample a random prompt, and press 'Send'",
-                    container=True,
-                    elem_id="input_box",
-                    lines = 15,
-                    interactive=False,      ## only support offline
-                )
-                
-                plausive_md = gr.Markdown("Geometry Plausibility: ", visible=False, elem_id="evaldim_markdown")
-                with gr.Row(elem_id="Geometry Plausibility"):
-                    plausive_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
-                    )
-                    plausive_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
-                    )
-                    plausive_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-                    plausive_bothbad_btn = gr.Button(
-                        value="👎  Both are bad", visible=False, interactive=False
-                    )
-
-                geo_md = gr.Markdown("Geometry Details: ", visible=False, elem_id="evaldim_markdown")
-                with gr.Row(elem_id="Geometry Details"):
-                    geo_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
-                    )
-                    geo_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
-                    )
-                    geo_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-                    geo_bothbad_btn = gr.Button(
-                        value="👎  Both are bad", visible=False, interactive=False
-                    )
-
-                text_md = gr.Markdown("Texture Quality: ", visible=False, elem_id="evaldim_markdown")
-                with gr.Row(elem_id="Texture Quality"):
-                    text_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
-                    )
-                    text_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
-                    )
-                    text_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-                    text_bothbad_btn = gr.Button(
-                        value="👎  Both are bad", visible=False, interactive=False
-                    )
-                
-                coherence_md = gr.Markdown("Geometry-Texture Coherency: ", visible=False, elem_id="evaldim_markdown")
-                with gr.Row(elem_id="Geometry-Texture Coherency"):
-                    coherence_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
-                    )
-                    coherence_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
-                    )
-                    coherence_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-                    coherence_bothbad_btn = gr.Button(
-                        value="👎  Both are bad", visible=False, interactive=False
-                    )
-
-                align_md = gr.Markdown("Semantic Alignment: ", visible=False, elem_id="evaldim_markdown")
-                with gr.Row(elem_id="Semantic Alignment"):
-                    align_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
-                    )
-                    align_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
-                    )
-                    align_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-                    align_bothbad_btn = gr.Button(
-                        value="👎  Both are bad", visible=False, interactive=False
-                        )
-            
-                sample_btn = gr.Button(value="🎲 Sample", variant="primary")
-
-            with gr.Column():
-                normal_left = gr.Image(width=512, label = "Normal A", show_download_button=True)
-                rgb_left = gr.Image(width=512, label = "RGB A", show_download_button=True)
-            with gr.Column():
-                normal_right = gr.Image(width=512, label = "Normal B", show_download_button=True,)
-                rgb_right = gr.Image(width=512, label = "RGB B", show_download_button=True,)
+        with gr.Column():
+            with gr.Row():
+                geo_left = gr.Image(label = "Geo A", interactive=False, show_download_button=True)
+                normal_left = gr.Image(label = "Normal A", interactive=False, show_download_button=True)
+                rgb_left = gr.Image(label = "RGB A", interactive=False, show_download_button=True)
+            with gr.Row():
+                geo_right = gr.Image(label = "Geo B", interactive=False, show_download_button=True)
+                normal_right = gr.Image(label = "Normal B", interactive=False, show_download_button=True)
+                rgb_right = gr.Image(label = "RGB B", interactive=False, show_download_button=True)
 
         with gr.Row():
-            with gr.Column():
-                gr.Markdown("", visible=False)
+            # with gr.Column():
+            #     gr.Markdown("", visible=False)
             with gr.Column():
                 model_selector_left =gr.Markdown("", visible=False)
             with gr.Column():
@@ -151,93 +76,93 @@ Find out who is the 🥇conditional image generation models! More models are goi
         with gr.Row():
             slow_warning = gr.Markdown("", elem_id="notice_markdown")
 
-        # with gr.Row(elem_id="Geometry Plausibility"):
-        #     plausive_md = gr.Markdown("Geometry Plausibility: ", visible=False, elem_id="evaldim_markdown")
-        #     plausive_leftvote_btn = gr.Button(
-        #         value="👈  A is better", visible=False, interactive=False
-        #     )
-        #     plausive_rightvote_btn = gr.Button(
-        #         value="👉  B is better", visible=False, interactive=False
-        #     )
-        #     plausive_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-        #     plausive_bothbad_btn = gr.Button(
-        #         value="👎  Both are bad", visible=False, interactive=False
-        #     )
-
-        # with gr.Row(elem_id="Geometry Details"):
-        #     geo_md = gr.Markdown("Geometry Details: ", visible=False, elem_id="evaldim_markdown")
-        #     geo_leftvote_btn = gr.Button(
-        #         value="👈  A is better", visible=False, interactive=False
-        #     )
-        #     geo_rightvote_btn = gr.Button(
-        #         value="👉  B is better", visible=False, interactive=False
-        #     )
-        #     geo_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-        #     geo_bothbad_btn = gr.Button(
-        #         value="👎  Both are bad", visible=False, interactive=False
-        #     )
-
-        # with gr.Row(elem_id="Texture Quality"):
-        #     text_md = gr.Markdown("Texture Quality: ", visible=False, elem_id="evaldim_markdown")
-        #     text_leftvote_btn = gr.Button(
-        #         value="👈  A is better", visible=False, interactive=False
-        #     )
-        #     text_rightvote_btn = gr.Button(
-        #         value="👉  B is better", visible=False, interactive=False
-        #     )
-        #     text_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-        #     text_bothbad_btn = gr.Button(
-        #         value="👎  Both are bad", visible=False, interactive=False
-        #     )
+        with gr.Row(elem_id="Geometry Plausibility"):
+            plausive_md = gr.Markdown("Geometry Plausibility: ", visible=False, elem_id="evaldim_markdown")
+            plausive_leftvote_btn = gr.Button(
+                value="👆  A is better", visible=False, interactive=False
+            )
+            plausive_rightvote_btn = gr.Button(
+                value="👇  B is better", visible=False, interactive=False
+            )
+            plausive_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
+            plausive_bothbad_btn = gr.Button(
+                value="👎  Both are bad", visible=False, interactive=False
+            )
+
+        with gr.Row(elem_id="Geometry Details"):
+            geo_md = gr.Markdown("Geometry Details: ", visible=False, elem_id="evaldim_markdown")
+            geo_leftvote_btn = gr.Button(
+                value="👆  A is better", visible=False, interactive=False
+            )
+            geo_rightvote_btn = gr.Button(
+                value="👇  B is better", visible=False, interactive=False
+            )
+            geo_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
+            geo_bothbad_btn = gr.Button(
+                value="👎  Both are bad", visible=False, interactive=False
+            )
+
+        with gr.Row(elem_id="Texture Quality"):
+            text_md = gr.Markdown("Texture Quality: ", visible=False, elem_id="evaldim_markdown")
+            text_leftvote_btn = gr.Button(
+                value="👆  A is better", visible=False, interactive=False
+            )
+            text_rightvote_btn = gr.Button(
+                value="👇  B is better", visible=False, interactive=False
+            )
+            text_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
+            text_bothbad_btn = gr.Button(
+                value="👎  Both are bad", visible=False, interactive=False
+            )
         
-        # with gr.Row(elem_id="Geometry-Texture Coherency"):
-        #     coherence_md = gr.Markdown("Geometry-Texture Coherency: ", visible=False, elem_id="evaldim_markdown")
-        #     coherence_leftvote_btn = gr.Button(
-        #         value="👈  A is better", visible=False, interactive=False
-        #     )
-        #     coherence_rightvote_btn = gr.Button(
-        #         value="👉  B is better", visible=False, interactive=False
-        #     )
-        #     coherence_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-        #     coherence_bothbad_btn = gr.Button(
-        #         value="👎  Both are bad", visible=False, interactive=False
-        #     )
-
-        # with gr.Row(elem_id="Semantic Alignment"):
-        #     align_md = gr.Markdown("Semantic Alignment: ", visible=False, elem_id="evaldim_markdown")
-        #     align_leftvote_btn = gr.Button(
-        #         value="👈  A is better", visible=False, interactive=False
-        #     )
-        #     align_rightvote_btn = gr.Button(
-        #         value="👉  B is better", visible=False, interactive=False
-        #     )
-        #     align_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-        #     align_bothbad_btn = gr.Button(
-        #         value="👎  Both are bad", visible=False, interactive=False
-            # )
-
-    # with gr.Row():
-    #     textbox = gr.Textbox(
-    #         show_label=False,
-    #         placeholder="👉 Enter your prompt or Sample a random prompt, and press ENTER",
-    #         container=True,
-    #         elem_id="input_box",
-    #         interactive=False,      ## only support offline
-    #     )
-    #     sample_btn = gr.Button(value="🎲 Sample", variant="primary", scale=0)
-    #     send_btn = gr.Button(value="📤 Send", variant="primary", scale=0)
+        with gr.Row(elem_id="Geometry-Texture Coherency"):
+            coherence_md = gr.Markdown("Geometry-Texture Coherency: ", visible=False, elem_id="evaldim_markdown")
+            coherence_leftvote_btn = gr.Button(
+                value="👆  A is better", visible=False, interactive=False
+            )
+            coherence_rightvote_btn = gr.Button(
+                value="👇  B is better", visible=False, interactive=False
+            )
+            coherence_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
+            coherence_bothbad_btn = gr.Button(
+                value="👎  Both are bad", visible=False, interactive=False
+            )
+
+        with gr.Row(elem_id="Semantic Alignment"):
+            align_md = gr.Markdown("Semantic Alignment: ", visible=False, elem_id="evaldim_markdown")
+            align_leftvote_btn = gr.Button(
+                value="👆  A is better", visible=False, interactive=False
+            )
+            align_rightvote_btn = gr.Button(
+                value="👇  B is better", visible=False, interactive=False
+            )
+            align_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
+            align_bothbad_btn = gr.Button(
+                value="👎  Both are bad", visible=False, interactive=False
+            )
+
+    with gr.Row():
+        textbox = gr.Textbox(
+            show_label=False,
+            placeholder="👉 Sample a random prompt, and press ENTER",
+            container=True,
+            elem_id="input_box",
+            interactive=False,      ## only support offline
+        )
+        sample_btn = gr.Button(value="🎲 Sample", variant="primary", scale=0)
+        send_btn = gr.Button(value="📤 Send", variant="primary", scale=0)
 
     with gr.Row():
         clear_btn = gr.Button(value="🗑️ Clear", interactive=False)
         regenerate_btn = gr.Button(value="🔄  Regenerate", interactive=False)
         share_btn = gr.Button(value="📷  Share")
-        send_btn = gr.Button(value="📤 Send", variant="primary")
+        # send_btn = gr.Button(value="📤 Send", variant="primary")
 
     gr.Markdown(acknowledgment_md, elem_id="ack_markdown")
 
     states = [state_0, state_1]
     model_selectors = [model_selector_left, model_selector_right]
-    results = [normal_left, rgb_left, normal_right, rgb_right]
+    results = [geo_left, normal_left, rgb_left, geo_right, normal_right, rgb_right]
 
     geo_btn_list = [geo_leftvote_btn, geo_rightvote_btn, geo_tie_btn, geo_bothbad_btn]
     text_btn_list = [text_leftvote_btn, text_rightvote_btn, text_tie_btn, text_bothbad_btn]
@@ -391,8 +316,8 @@ Find out who is the 🥇conditional image generation models! More models are goi
             gr.Markdown(model_description_md, elem_id="model_description_markdown")
             
         with gr.Row():
-            with gr.Column():
-                gr.Markdown("Select Players:", elem_id="evaldim_markdown")
+            # with gr.Column():
+            #     gr.Markdown("Select Players:", elem_id="evaldim_markdown")
             with gr.Column():
                 model_selector_left = gr.Dropdown(
                     choices=model_list,
@@ -410,105 +335,105 @@ Find out who is the 🥇conditional image generation models! More models are goi
                     container=False,
                     )
 
-        with gr.Row():
-            with gr.Column():
-                textbox = gr.Textbox(
-                    show_label=False,
-                    placeholder="👉 Enter your prompt or Sample a random prompt, and press 'Send'",
-                    container=True,
-                    elem_id="input_box",
-                    lines=15,
-                    interactive=False,   ## only support offline
-                )
-                        
-                plausive_md = gr.Markdown("Geometry Plausibility: ", visible=False, elem_id="evaldim_markdown")
-                with gr.Row(elem_id="Geometry Plausibility"):
-                    plausive_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
-                    )
-                    plausive_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
-                    )
-                    plausive_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-                    plausive_bothbad_btn = gr.Button(
-                        value="👎  Both are bad", visible=False, interactive=False
-                    )
-                
-                geo_md = gr.Markdown("Geometry Details: ", visible=False, elem_id="evaldim_markdown")
-                with gr.Row(elem_id="Geometry Details"):
-                    geo_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
-                    )
-                    geo_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
-                    )
-                    geo_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-                    geo_bothbad_btn = gr.Button(
-                        value="👎  Both are bad", visible=False, interactive=False
-                    )
+        with gr.Column():
+            with gr.Row():
+                geo_left = gr.Image(label = "Geo A", interactive=False, show_download_button=True)
+                normal_left = gr.Image(label = "Normal A", interactive=False, show_download_button=True)
+                rgb_left = gr.Image(label = "RGB A", interactive=False, show_download_button=True)
+            with gr.Row():
+                geo_right = gr.Image(label = "Geo B", interactive=False, show_download_button=True)
+                normal_right = gr.Image(label = "Normal B", interactive=False, show_download_button=True)
+                rgb_right = gr.Image(label = "RGB B", interactive=False, show_download_button=True)
 
-                text_md = gr.Markdown("Texture Quality: ", visible=False, elem_id="evaldim_markdown")
-                with gr.Row(elem_id="Texture Quality"):
-                    text_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
-                    )
-                    text_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
-                    )
-                    text_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-                    text_bothbad_btn = gr.Button(
-                        value="👎  Both are bad", visible=False, interactive=False
-                    )
-                
-                coherence_md = gr.Markdown("Geometry-Texture Coherency: ", visible=False, elem_id="evaldim_markdown")
-                with gr.Row(elem_id="Geometry-Texture Coherency"):
-                    coherence_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
-                    )
-                    coherence_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
-                    )
-                    coherence_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-                    coherence_bothbad_btn = gr.Button(
-                        value="👎  Both are bad", visible=False, interactive=False
-                    )
+        with gr.Row():
+            slow_warning = gr.Markdown("", elem_id="notice_markdown")
 
-                align_md = gr.Markdown("Semantic Alignment: ", visible=False, elem_id="evaldim_markdown")
-                with gr.Row(elem_id="Semantic Alignment"):
-                    align_leftvote_btn = gr.Button(
-                        value="👈  A is better", visible=False, interactive=False
-                    )
-                    align_rightvote_btn = gr.Button(
-                        value="👉  B is better", visible=False, interactive=False
-                    )
-                    align_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
-                    align_bothbad_btn = gr.Button(
-                        value="👎  Both are bad", visible=False, interactive=False
-                    )
+        with gr.Row(elem_id="Geometry Plausibility"):
+            plausive_md = gr.Markdown("Geometry Plausibility: ", visible=False, elem_id="evaldim_markdown")
+            plausive_leftvote_btn = gr.Button(
+                value="👆  A is better", visible=False, interactive=False
+            )
+            plausive_rightvote_btn = gr.Button(
+                value="👇  B is better", visible=False, interactive=False
+            )
+            plausive_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
+            plausive_bothbad_btn = gr.Button(
+                value="👎  Both are bad", visible=False, interactive=False
+            )
+        
+        with gr.Row(elem_id="Geometry Details"):
+            geo_md = gr.Markdown("Geometry Details: ", visible=False, elem_id="evaldim_markdown")
+            geo_leftvote_btn = gr.Button(
+                value="👆  A is better", visible=False, interactive=False
+            )
+            geo_rightvote_btn = gr.Button(
+                value="👇  B is better", visible=False, interactive=False
+            )
+            geo_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
+            geo_bothbad_btn = gr.Button(
+                value="👎  Both are bad", visible=False, interactive=False
+            )
 
-                sample_btn = gr.Button(value="🎲 Sample", variant="primary")
+        with gr.Row(elem_id="Texture Quality"):
+            text_md = gr.Markdown("Texture Quality: ", visible=False, elem_id="evaldim_markdown")
+            text_leftvote_btn = gr.Button(
+                value="👆  A is better", visible=False, interactive=False
+            )
+            text_rightvote_btn = gr.Button(
+                value="👇  B is better", visible=False, interactive=False
+            )
+            text_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
+            text_bothbad_btn = gr.Button(
+                value="👎  Both are bad", visible=False, interactive=False
+            )
+        
+        with gr.Row(elem_id="Geometry-Texture Coherency"):
+            coherence_md = gr.Markdown("Geometry-Texture Coherency: ", visible=False, elem_id="evaldim_markdown")
+            coherence_leftvote_btn = gr.Button(
+                value="👆  A is better", visible=False, interactive=False
+            )
+            coherence_rightvote_btn = gr.Button(
+                value="👇  B is better", visible=False, interactive=False
+            )
+            coherence_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
+            coherence_bothbad_btn = gr.Button(
+                value="👎  Both are bad", visible=False, interactive=False
+            )
 
-            with gr.Column():
-                normal_left = gr.Image(width=512, label = "Normal A", show_download_button=True)
-                rgb_left = gr.Image(width=512, label = "RGB A", show_download_button=True)
-            with gr.Column():
-                normal_right = gr.Image(width=512, label = "Normal B", show_download_button=True,)
-                rgb_right = gr.Image(width=512, label = "RGB B", show_download_button=True,)
+        with gr.Row(elem_id="Semantic Alignment"):
+            align_md = gr.Markdown("Semantic Alignment: ", visible=False, elem_id="evaldim_markdown")
+            align_leftvote_btn = gr.Button(
+                value="👆  A is better", visible=False, interactive=False
+            )
+            align_rightvote_btn = gr.Button(
+                value="👇  B is better", visible=False, interactive=False
+            )
+            align_tie_btn = gr.Button(value="🤝  Tie", visible=False, interactive=False)
+            align_bothbad_btn = gr.Button(
+                value="👎  Both are bad", visible=False, interactive=False
+            )
 
-        with gr.Row():
-            slow_warning = gr.Markdown("", elem_id="notice_markdown")
+    with gr.Row():
+        textbox = gr.Textbox(
+            show_label=False,
+            placeholder="👉 Sample a random prompt, sand press 'Send'",
+            container=True,
+            elem_id="input_box",
+            interactive=False,   ## only support offline
+        )   
+        sample_btn = gr.Button(value="🎲 Sample", variant="primary", scale=0)
+        send_btn = gr.Button(value="📤 Send", variant="primary", scale=0)
 
     with gr.Row():
         clear_btn = gr.Button(value="🗑️ Clear", interactive=False)
         regenerate_btn = gr.Button(value="🔄  Regenerate", interactive=False)
         share_btn = gr.Button(value="📷  Share")
-        send_btn = gr.Button(value="📤 Send", variant="primary")
 
     gr.Markdown(acknowledgment_md, elem_id="ack_markdown")
 
     states = [state_0, state_1]
     model_selectors = [model_selector_left, model_selector_right]
-    results = [normal_left, rgb_left, normal_right, rgb_right]
+    results = [geo_left, normal_left, rgb_left, geo_right, normal_right, rgb_right]
 
     geo_btn_list = [geo_leftvote_btn, geo_rightvote_btn, geo_tie_btn, geo_bothbad_btn]
     text_btn_list = [text_leftvote_btn, text_rightvote_btn, text_tie_btn, text_bothbad_btn]
@@ -689,18 +614,9 @@ def build_t2s_ui_single_model(models):
         )
 
     with gr.Row():
-        with gr.Column():
-            textbox = gr.Textbox(
-                show_label=False,
-                placeholder="👉 Enter your prompt or Sample a random prompt, and press 'Send'",
-                container=True,
-                elem_id="input_box",
-                lines=15,
-                interactive=False,   ## only support offline
-            )
-            sample_btn = gr.Button(value="🎲 Sample", variant="primary")
-        normal = gr.Image(width=512, label = "Normal", show_download_button=True)
-        rgb = gr.Image(width=512, label = "RGB", show_download_button=True,)
+        geo = gr.Image(label = "Geo", interactive=False, show_download_button=True)
+        normal = gr.Image(label = "Normal", interactive=False, show_download_button=True)
+        rgb = gr.Image(label = "RGB", interactive=False, show_download_button=True)
     
     with gr.Row(elem_id="Geometry Plausibility"):
         plausive_md = gr.Markdown("Geometry Plausibility: ", elem_id="evaldim_markdown")
@@ -732,10 +648,20 @@ def build_t2s_ui_single_model(models):
         align_downvote_btn = gr.Button(value="👎  Downvote", interactive=False)
         align_flag_btn = gr.Button(value="⚠️  Flag", interactive=False)
 
+    with gr.Row():
+        textbox = gr.Textbox(
+            show_label=False,
+            placeholder="👉 Sample a random prompt, and press 'Send'",
+            container=True,
+            elem_id="input_box",
+            interactive=False,   ## only support offline
+        )
+        sample_btn = gr.Button(value="🎲 Sample", variant="primary", scale=0)
+        send_btn = gr.Button(value="📤 Send", variant="primary",scale=0)
+
     with gr.Row():
         clear_btn = gr.Button(value="🗑️ Clear", interactive=False)
         regenerate_btn = gr.Button(value="🔄  Regenerate", interactive=False)
-        send_btn = gr.Button(value="📤 Send", variant="primary")
 
     gr.Markdown(acknowledgment_md, elem_id="ack_markdown")
 
@@ -771,7 +697,7 @@ def build_t2s_ui_single_model(models):
     model_selector.change(
         clear_t2s_history, 
         None, 
-        [state, textbox, normal, rgb], 
+        [state, textbox, geo, normal, rgb], 
         api_name="clear_history_single",
         show_progress="full"
     ).then(
@@ -788,17 +714,17 @@ def build_t2s_ui_single_model(models):
     ).then(
         reset_state,
         state,
-        [state, normal, rgb] + vote_btn_list + [regenerate_btn, clear_btn]
+        [state, geo, normal, rgb] + vote_btn_list + [regenerate_btn, clear_btn]
     )
 
     textbox.submit(
         reset_state,
         state,
-        [state, normal, rgb] + vote_btn_list + [regenerate_btn, clear_btn]
+        [state, geo, normal, rgb] + vote_btn_list + [regenerate_btn, clear_btn]
     ).then(
         gen_func,
         [state, textbox, model_selector],
-        [state, normal, rgb],
+        [state, geo, normal, rgb],
         api_name="submit_btn_single",
         show_progress = "full"
     ).then(
@@ -810,11 +736,11 @@ def build_t2s_ui_single_model(models):
     send_btn.click(
         reset_state,
         state,
-        [state, normal, rgb] + vote_btn_list + [regenerate_btn, clear_btn]
+        [state, geo, normal, rgb] + vote_btn_list + [regenerate_btn, clear_btn]
     ).then(
         gen_func,
         [state, textbox, model_selector],
-        [state, normal, rgb],
+        [state, geo, normal, rgb],
         api_name="send_btn_single",
         show_progress = "full"
     ).then(
@@ -826,7 +752,7 @@ def build_t2s_ui_single_model(models):
     clear_btn.click(
         clear_t2s_history, 
         None, 
-        [state, textbox, normal, rgb], 
+        [state, textbox, geo, normal, rgb], 
         api_name="clear_history_single",
         show_progress="full"
     ).then(
@@ -838,11 +764,11 @@ def build_t2s_ui_single_model(models):
     regenerate_btn.click(
         reset_state,
         state,
-        [state, normal, rgb] + vote_btn_list + [regenerate_btn, clear_btn]
+        [state, geo, normal, rgb] + vote_btn_list + [regenerate_btn, clear_btn]
     ).then(
         gen_func,
         [state, textbox, model_selector],
-        [state, normal, rgb],
+        [state, geo, normal, rgb],
         api_name="regenerate_btn_single",
         show_progress = "full"
     ).then(

serve/inference.py

@@ -15,7 +15,7 @@ class State:
     def __init__(self, 
                  model_name, i2s_mode=False, offline=False, 
                  prompt=None, image=None, offline_idx=None, 
-                 normal_video=None , rgb_video=None,
+                 normal_video=None , rgb_video=None, geo_video=None,
                  evaluted_dims=0):
         self.conv_id = uuid.uuid4().hex
         self.model_name = model_name
@@ -28,6 +28,7 @@ class State:
         # self.output = None
         self.normal_video = normal_video
         self.rgb_video = rgb_video
+        self.geo_video = geo_video
 
         self.evaluted_dims = evaluted_dims
 
@@ -180,8 +181,9 @@ def generate_t2s(gen_func, render_func,
         # rgb_video = os.path.join(OFFLINE_DIR, "text2shape", model_name, "rgb", f"{state.offline_idx}.mp4")
         
         state.normal_video = videos['normal']
-        state.rgb_video = videos['rgb']        
-        yield state, videos['normal'], videos['rgb']
+        state.rgb_video = videos['rgb']
+        state.geo_video = videos['geo']         
+        yield state, videos['geo'], videos['normal'], videos['rgb']
 
         finish_tstamp = time.time()
         # logger.info(f"===output===: {output}")
@@ -274,9 +276,11 @@ def generate_t2s_multi(gen_func, render_func,
         # normal_video_1 = os.path.join(OFFLINE_DIR, "text2shape", model_name_1, "normal", f"{state_1.offline_idx}.mp4")
         # rgb_video_1 = os.path.join(OFFLINE_DIR, "text2shape", model_name_1, "rgb", f"{state_1.offline_idx}.mp4")
         
-        state_0.normal_video, state_0.rgb_video = videos_0['normal'], videos_0['rgb']
-        state_1.normal_video, state_1.rgb_video = videos_1['normal'], videos_1['rgb']
-        yield state_0, state_1,videos_0['normal'], videos_0['rgb'], videos_1['normal'], videos_1['rgb']
+        state_0.normal_video, state_0.rgb_video, state_0.geo_video = videos_0['normal'], videos_0['rgb'], videos_0['geo']
+        state_1.normal_video, state_1.rgb_video, state_1.geo_video = videos_1['normal'], videos_1['rgb'], videos_1['geo']
+        yield state_0, state_1, \
+            videos_0['geo'], videos_0['normal'], videos_0['rgb'], \
+            videos_1['geo'], videos_1['normal'], videos_1['rgb']
 
         finish_tstamp = time.time()
         # logger.info(f"===output===: {output}")
@@ -308,10 +312,12 @@ def generate_t2s_multi(gen_func, render_func,
         videos_0, videos_1 = render_func(shape_0, model_name_0, shape_1, model_name_1)
         finish_time = time.time()
         render_time = finish_time - start_time - generate_time
-        
-        state_0.normal_video, state_0.rgb_video = videos_0['normal'], videos_0['rgb']
-        state_1.normal_video, state_1.rgb_video = videos_1['normal'], videos_1['rgb']
-        yield state_0, state_1, videos_0['normal'], videos_0['rgb'], videos_1['normal'], videos_1['rgb']
+
+        state_0.normal_video, state_0.rgb_video, state_0.geo_video = videos_0['normal'], videos_0['rgb'], videos_0['geo']
+        state_1.normal_video, state_1.rgb_video, state_1.geo_video = videos_1['normal'], videos_1['rgb'], videos_1['geo']
+        yield state_0, state_1, \
+            videos_0['geo'], videos_0['normal'], videos_0['rgb'], \
+            videos_1['geo'], videos_1['normal'], videos_1['rgb']
 
         finish_tstamp = time.time()
         # logger.info(f"===output===: {output}")
@@ -392,9 +398,11 @@ def generate_t2s_multi_annoy(gen_func, render_func,
         # rgb_video_1 = os.path.join(OFFLINE_DIR, "text2shape", model_name_1, "rgb", f"{state_1.offline_idx}.mp4")
         
         state_0.model_name, state_1.model_name = model_name_0, model_name_1
-        state_0.normal_video, state_0.rgb_video = videos_0['normal'], videos_0['rgb']
-        state_1.normal_video, state_1.rgb_video = videos_1['normal'], videos_1['rgb']
-        yield state_0, state_1, videos_0['normal'], videos_0['rgb'], videos_1['normal'], videos_1['rgb'], \
+        state_0.normal_video, state_0.rgb_video, state_0.geo_video = videos_0['normal'], videos_0['rgb'], videos_0['geo']
+        state_1.normal_video, state_1.rgb_video, state_1.geo_video = videos_1['normal'], videos_1['rgb'], videos_1['geo']
+        yield state_0, state_1, \
+            videos_0['geo'], videos_0['normal'], videos_0['rgb'], \
+            videos_1['geo'], videos_1['normal'], videos_1['rgb'], \
             gr.Markdown(f"### Model A: {model_name_0}"), gr.Markdown(f"### Model B: {model_name_1}")
 
         finish_tstamp = time.time()
@@ -429,9 +437,11 @@ def generate_t2s_multi_annoy(gen_func, render_func,
         render_time = finish_time - start_time - generate_time
         
         state_0.model_name, state_1.model_name = model_name_0, model_name_1
-        state_0.normal_video, state_0.rgb_video = videos_0['normal'], videos_0['rgb']
-        state_1.normal_video, state_1.rgb_video = videos_1['normal'], videos_1['rgb']
-        yield state_0, state_1, videos_0[0], videos_0[1], videos_1[0], videos_1[1], \
+        state_0.normal_video, state_0.rgb_video, state_0.geo_video = videos_0['normal'], videos_0['rgb'], videos_0['geo']
+        state_1.normal_video, state_1.rgb_video, state_1.geo_video = videos_1['normal'], videos_1['rgb'], videos_1['geo']
+        yield state_0, state_1, \
+            videos_0['geo'], videos_0['normal'], videos_0['rgb'], \
+            videos_1['geo'], videos_1['normal'], videos_1['rgb'], \
             gr.Markdown(f"### Model A: {model_name_0}"), gr.Markdown(f"### Model B: {model_name_1}")
 
         finish_tstamp = time.time()
@@ -500,7 +510,8 @@ def generate_i2s(gen_func, render_func, state, image, model_name, request: gr.Re
         
         state.normal_video = videos['normal']
         state.rgb_video = videos['rgb']
-        yield state, videos['normal'], videos['rgb']
+        state.geo_video = videos['geo']
+        yield state, videos['geo'], videos['normal'], videos['rgb']
 
         finish_tstamp = time.time()
         # logger.info(f"===output===: {output}")
@@ -589,9 +600,11 @@ def generate_i2s_multi(gen_func, render_func,
         # normal_video_1 = os.path.join(OFFLINE_DIR, "image2shape", model_name_1, "normal", f"{state_1.offline_idx}.mp4")
         # rgb_video_1 = os.path.join(OFFLINE_DIR, "image2shape", model_name_1, "rgb", f"{state_1.offline_idx}.mp4")
         
-        state_0.normal_video, state_0.rgb_video = videos_0['normal'], videos_0['rgb']
-        state_1.normal_video, state_1.rgb_video = videos_1['normal'], videos_1['rgb']
-        yield state_0, state_1, videos_0['normal'], videos_0['rgb'], videos_1['normal'], videos_1['rgb'], \
+        state_0.normal_video, state_0.rgb_video, state_0.geo_video = videos_0['normal'], videos_0['rgb'], videos_0['geo']
+        state_1.normal_video, state_1.rgb_video, state_1.geo_video = videos_1['normal'], videos_1['rgb'], videos_1['geo']
+        yield state_0, state_1, \
+            videos_0['geo'], videos_0['normal'], videos_0['rgb'], \
+            videos_1['geo'], videos_1['normal'], videos_1['rgb'], \
             gr.Markdown(f"### Model A: {model_name_0}"), gr.Markdown(f"### Model B: {model_name_1}")
 
         finish_tstamp = time.time()
@@ -625,9 +638,11 @@ def generate_i2s_multi(gen_func, render_func,
         finish_time = time.time()
         render_time = finish_time - start_time - generate_time
 
-        state_0.normal_video, state_0.rgb_video = videos_0['normal'], videos_0['rgb']
-        state_1.normal_video, state_1.rgb_video = videos_1['normal'], videos_1['rgb']
-        yield state_0, state_1, videos_0['normal'], videos_0['rgb'], videos_1['normal'], videos_1['rgb']
+        state_0.normal_video, state_0.rgb_video, state_0.geo_video = videos_0['normal'], videos_0['rgb'], videos_0['geo']
+        state_1.normal_video, state_1.rgb_video, state_1.geo_video = videos_1['normal'], videos_1['rgb'], videos_1['geo']
+        yield state_0, state_1, \
+            videos_0['geo'], videos_0['normal'], videos_0['rgb'], \
+            videos_1['geo'], videos_1['normal'], videos_1['rgb']
 
         finish_tstamp = time.time()
         # logger.info(f"===output===: {output}")
@@ -704,9 +719,11 @@ def generate_i2s_multi_annoy(gen_func, render_func,
         # normal_video_1 = os.path.join(OFFLINE_DIR, "image2shape", model_name_1, "normal", f"{state_1.offline_idx}.mp4")
         # rgb_video_1 = os.path.join(OFFLINE_DIR, "image2shape", model_name_1, "rgb", f"{state_1.offline_idx}.mp4")
         state_0.model_name, state_1.model_name = model_name_0, model_name_1
-        state_0.normal_video, state_0.rgb_video = videos_0['normal'], videos_0['rgb']
-        state_1.normal_video, state_1.rgb_video = videos_1['normal'], videos_1['rgb']
-        yield state_0, state_1, videos_0['normal'], videos_0['rgb'], videos_1['normal'], videos_1['rgb'], \
+        state_0.normal_video, state_0.rgb_video, state_0.geo_video = videos_0['normal'], videos_0['rgb'], videos_0['geo']
+        state_1.normal_video, state_1.rgb_video, state_1.geo_video = videos_1['normal'], videos_1['rgb'], videos_1['geo']
+        yield state_0, state_1, \
+            videos_0['geo'], videos_0['normal'], videos_0['rgb'], \
+            videos_1['geo'], videos_1['normal'], videos_1['rgb'], \
             gr.Markdown(f"### Model A: {model_name_0}"), gr.Markdown(f"### Model B: {model_name_1}")
 
         finish_tstamp = time.time()
@@ -740,9 +757,11 @@ def generate_i2s_multi_annoy(gen_func, render_func,
         finish_time = time.time()
         render_time = finish_time - start_time - generate_time
 
-        state_0.normal_video, state_0.rgb_video = videos_0['normal'], videos_0['rgb']
-        state_1.normal_video, state_1.rgb_video = videos_1['normal'], videos_1['rgb']
-        yield state_0, state_1,  videos_0['normal'], videos_0['rgb'], videos_1['normal'], videos_1['rgb'], \
+        state_0.normal_video, state_0.rgb_video, state_0.geo_video = videos_0['normal'], videos_0['rgb'], videos_0['geo']
+        state_1.normal_video, state_1.rgb_video, state_1.geo_video = videos_1['normal'], videos_1['rgb'], videos_1['geo']
+        yield state_0, state_1, \
+            videos_0['geo'], videos_0['normal'], videos_0['rgb'], \
+            videos_1['geo'], videos_1['normal'], videos_1['rgb'], \
             gr.Markdown(f"### Model A: {model_name_0}"), gr.Markdown(f"### Model B: {model_name_1}")
 
         finish_tstamp = time.time()

serve/utils.py

@@ -20,6 +20,7 @@ no_change_btn = gr.update(value="No Change", interactive=True, visible=True)
 def build_about():
     about_markdown = f"""
 # About Us
+Only offline services are available currently, online services are coming soon!
 Supported by Shanghai AI Laboratory
 
 ## Contributors:
@@ -120,45 +121,45 @@ def disable_buttons():
 def reset_state(state):
     state.normal_video, state.rgb_video = None, None
     state.evaluted_dims = 0
-    return (state, None, None) + tuple(gr.update(interactive=False) for _ in range(EVALUATE_DIMS*3 + 2))
+    return (state, None, None, None) + tuple(gr.update(interactive=False) for _ in range(EVALUATE_DIMS*3 + 2))
 
 def reset_states_side_by_side(state_0, state_1):
-    state_0.normal_video, state_0.rgb_video = None, None
-    state_1.normal_video, state_1.rgb_video = None, None
+    state_0.normal_video, state_0.rgb_video, state_0.geo_video = None, None, None
+    state_1.normal_video, state_1.rgb_video, state_1.geo_video = None, None, None
     state_0.evaluted_dims, state_1.evaluted_dims = 0, 0
     return (state_0, state_1) \
-        + (None,) * 4\
+        + (None,) * 6\
         + tuple(gr.update(visible=(i>=EVALUATE_DIMS*4), interactive=False) for i in range(EVALUATE_DIMS*4 + 2)) \
         + tuple(gr.update(visible=False) for _ in range(EVALUATE_DIMS))
 
 def reset_states_side_by_side_anony(state_0, state_1):
     state_0.model_name, state_1.model_name = "", ""
-    state_0.normal_video, state_0.rgb_video = None, None
-    state_1.normal_video, state_1.rgb_video = None, None
+    state_0.normal_video, state_0.rgb_video, state_0.geo_video = None, None, None
+    state_1.normal_video, state_1.rgb_video, state_1.geo_video = None, None, None
     state_0.evaluted_dims, state_1.evaluted_dims = 0, 0
     return (state_0, state_1) \
         + (gr.Markdown("", visible=False), gr.Markdown("", visible=False))\
-        + (None,) * 4 \
+        + (None,) * 6 \
         + tuple(gr.update(visible=(i>=EVALUATE_DIMS*4), interactive=False) for i in range(EVALUATE_DIMS*4 + 2)) \
         + tuple(gr.update(visible=False) for _ in range(EVALUATE_DIMS))
 
 def clear_t2s_history():
-    return None, "", None, None
+    return None, "", None, None, None
    
 def clear_t2s_history_side_by_side():
-    return [None] * 2 + [""] + [None] * 4
+    return [None] * 2 + [""] + [None] * 6
 
 def clear_t2s_history_side_by_side_anony():
-    return [None] * 2 + [""] + [None] * 4 + [gr.Markdown("", visible=False), gr.Markdown("", visible=False)]
+    return [None] * 2 + [""] + [None] * 6 + [gr.Markdown("", visible=False), gr.Markdown("", visible=False)]
 
 def clear_i2s_history():
-    return None, None, None, None
+    return None, None, None, None, None
    
 def clear_i2s_history_side_by_side():
-    return [None] * 2 + [None] + [None] * 4
+    return [None] * 2 + [None] + [None] * 6
 
 def clear_i2s_history_side_by_side_anony():
-    return [None] * 2 + [None] + [None] * 4 + [gr.Markdown("", visible=False), gr.Markdown("", visible=False)]
+    return [None] * 2 + [None] + [None] * 6 + [gr.Markdown("", visible=False), gr.Markdown("", visible=False)]
 
 def get_ip(request: gr.Request):
     if request: