app.py

import gradio as gr
import torch
import numpy as np
from huggingface_hub import hf_hub_download
from depth_anything_v2.dpt import DepthAnythingV2

DEVICE = 'cuda' if torch.cuda.is_available() else 'cpu'
model_configs = {
    'vitl': {'encoder': 'vitl', 'features': 256, 'out_channels': [256, 512, 1024, 1024]},
}
encoder = 'vitl'
model = DepthAnythingV2(**model_configs[encoder])
model_path = hf_hub_download(
    repo_id="depth-anything/Depth-Anything-V2-Large",
    filename=f"depth_anything_v2_{encoder}.pth",
    repo_type="model"
)
state_dict = torch.load(model_path, map_location="cpu")
model.load_state_dict(state_dict)
model = model.to(DEVICE).eval()

def infer(img):
    with torch.no_grad():
        depth = model.infer_image(img[:, :, ::-1])  # BGR to RGB if needed
        # Normalize to 0-255 and convert to uint8
        depth_norm = (depth - depth.min()) / (depth.max() - depth.min()) * 255.0
        return depth_norm.astype(np.uint8)

iface = gr.Interface(fn=infer, inputs=gr.Image(type="numpy"), outputs=gr.Image())
iface.launch()