lianghsun commited on
Commit
0a08264
·
1 Parent(s): cdf2972

First commit

Browse files
.gitignore ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ /src
2
+ *.py
README.md CHANGED
@@ -1,3 +1,41 @@
1
- ---
2
- license: cc-by-nc-sa-4.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-nc-sa-4.0
3
+ annotations_creators:
4
+ - machine-generated
5
+ language:
6
+ - en
7
+ - zh
8
+ language_creators:
9
+ - other
10
+ multilinguality:
11
+ - multilingual
12
+ pretty_name: 'Pokémon BLIP captions'
13
+ size_categories:
14
+ - n<1K
15
+ source_datasets:
16
+ - svjack/pokemon-blip-captions-en-zh
17
+ tags: []
18
+ task_categories:
19
+ - text-to-image
20
+ task_ids: []
21
+ ---
22
+
23
+ # Data Card for Pokémon BLIP captions with English and Traditional Chinese
24
+
25
+ [![Hugging Face Space](https://img.shields.io/badge/fork-huggingface-blue?style=social)](https://huggingface.co/datasets/svjack/pokemon-blip-captions-en-zh)
26
+
27
+ ## Dataset Summary
28
+
29
+ 本資料集用於訓練寶可夢文本生成圖像模型,**新增**了繁體中文欄位:`zh_tw_text`。並**移除**原先的簡體中文欄位:`zh_tw`。
30
+
31
+ [BLIP](https://arxiv.org/abs/2201.12086) 為寶可夢圖像生成了標註(captions),這些圖像來自於「[Towards Faster and Stabilized GAN Training for High-fidelity Few-shot Image Synthesis (FastGAN)](https://openreview.net/forum?id=1Fqg133qRaI)」中引入的少量樣本寶可夢資料集。原始圖像是從 [FastGAN-pytorch](https://github.com/odegeasslbc/FastGAN-pytorch?tab=readme-ov-file) 獲得的,並使用預訓練的 BLIP 模型生成標註。
32
+
33
+ 對於每一行資料,資料集包含`圖像`、`en_text`(英文標註)和 `zh_tw_text`(繁體中文標註)欄位。圖像為不同尺寸的 PIL jpeg,文本為相應的標註文本。僅提供訓練集分割(train split),並未提供測試集分割(eval split)。
34
+
35
+ [原始資料集](https://huggingface.co/datasets/svjack/pokemon-blip-captions-en-zh)內的簡體中文標註是透過 Deepl 翻譯生成的,然而在此延伸版本的資料集內的繁體中文翻譯是透過 [opencc-python-reimplemented](https://pypi.org/project/opencc-python-reimplemented/) 套件完成的。
36
+
37
+ ## Languages
38
+ 繁體中文。
39
+
40
+ ## License
41
+ Creative Commons Attribution Non Commercial Share Alike 4.0
data/train-00000-of-00001-78e564002aa9c8f0.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:225d64d75d91604c2c0880d5a55ecee77340c4b3449f4873f826060bb0fc3a0e
3
+ size 99688153
dataset_infos.json ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "lianghsun--pokemon-blip-captions-en-zh_tw": {
3
+ "description": "This dataset is a fork from the original svjack/pokemon-blip-captions-en-zh. It contains captions generated for Pokémon images from the Few Shot Pokémon dataset introduced by 'Towards Faster and Stabilized GAN Training for High-fidelity Few-shot Image Synthesis (FastGAN)'. The original images were captioned using the pre-trained BLIP model, with English captions in the 'en_text' field and Chinese captions in the 'zh_tw_text' field, which were converted to Traditional Chinese using the opencc-python-reimplemented package.",
4
+ "citation": "If you use this dataset, please cite the original work from svjack and FastGAN: \n\n@article{li2021fastgan,\n title={Towards Faster and Stabilized GAN Training for High-fidelity Few-shot Image Synthesis},\n author={Bingchen Liu, Yizhe Zhu, Kunpeng Song, Ahmed Elgammal},\n journal={ICLR 2021},\n year={2021}\n}",
5
+ "homepage": "https://huggingface.co/datasets/lianghsun/pokemon-blip-captions-en-zh_tw",
6
+ "license": "Creative Commons Attribution Non Commercial Share Alike 4.0",
7
+ "features": {
8
+ "image": {
9
+ "decode": true,
10
+ "id": null,
11
+ "_type": "Image"
12
+ },
13
+ "en_text": {
14
+ "dtype": "string",
15
+ "id": null,
16
+ "_type": "Value"
17
+ },
18
+ "zh_tw_text": {
19
+ "dtype": "string",
20
+ "id": null,
21
+ "_type": "Value"
22
+ }
23
+ },
24
+ "post_processed": "Removed zh_text and added zh_tw_text using opencc-python-reimplemented to convert to Traditional Chinese. Additionally, replaced '小精靈' with '寶可夢'.",
25
+ "supervised_keys": null,
26
+ "task_templates": null,
27
+ "builder_name": "imagefolder",
28
+ "config_name": "pokemon",
29
+ "version": {
30
+ "version_str": "1.0.0",
31
+ "description": "Initial release with Traditional Chinese translations",
32
+ "major": 1,
33
+ "minor": 0,
34
+ "patch": 0
35
+ },
36
+ "splits": {
37
+ "train": {
38
+ "name": "train",
39
+ "num_bytes": 119450251.0,
40
+ "num_examples": 833,
41
+ "dataset_name": "pokemon-blip-captions-en-zh_tw"
42
+ }
43
+ },
44
+ "download_checksums": null,
45
+ "download_size": 99686892,
46
+ "post_processing_size": null,
47
+ "dataset_size": 119450251.0,
48
+ "size_in_bytes": 219137143.0
49
+ }
50
+ }