README.md

---
license: apache-2.0
# inference: false
pipeline_tag: zero-shot-image-classification

# inference:
#   parameters:
tags:
- clip
- zh
- image-text
---

# Model Details

This model is a Chinese CLIP model trained on [Noah-Wukong Dataset](https://wukong-dataset.github.io/wukong-dataset/), which contains about 100M Chinese image-text pairs. We use ViT-B-32 from [openAI](https://github.com/openai/CLIP) as image encoder and Chinese pre-trained language model  [chinese-roberta-wwm](https://huggingface.co/hfl/chinese-roberta-wwm-ext) as text encoder. We freeze the image encoder and only finetune the text encoder. The model was trained for 20 epochs and it takes about 10 days with 8 A100 GPUs.

# Taiyi (太乙)
Taiyi models are a branch of the Fengshenbang (封神榜) series of models. The models in Taiyi are pre-trained with multimodal pre-training strategies. We will release more image-text model trained on Chinese dataset and benefit the Chinese community.



# Usage

```python3
from PIL import Image
import requests
import clip
import torch
from transformers import BertForSequenceClassification, BertConfig, BertTokenizer
import numpy as np

# 加载TaiYi 中文 text encoder
text_tokenizer = BertTokenizer.from_pretrained("IDEA-CCNL/TaiYi-CLIP-Roberta-Chinese")
text_encoder = BertForSequenceClassification.from_pretrained("IDEA-CCNL/TaiYi-CLIP-Roberta-Chinese").eval()
text = text_tokenizer(["一只猫", "一只狗",'两只猫', '两只老虎','一只老虎'], return_tensors='pt', padding=True)['input_ids']

# 加载CLIP的image encoder
url = "http://images.cocodataset.org/val2017/000000039769.jpg"
clip_model, preprocess = clip.load("ViT-B/32", device='cpu')
image = preprocess(Image.open(requests.get(url, stream=True).raw)).unsqueeze(0)

with torch.no_grad():
    image_features = clip_model.encode_image(image)
    text_features = text_encoder(text).logits
    # 归一化
    image_features = image_features / image_features.norm(dim=1, keepdim=True)
    text_features = text_features / text_features.norm(dim=1, keepdim=True)
    # 计算余弦相似度 logit_scale是尺度系数
    logit_scale = clip_model.logit_scale.exp()
    logits_per_image = logit_scale * image_features @ text_features.t()
    logits_per_text = logits_per_image.t()
    probs = logits_per_image.softmax(dim=-1).cpu().numpy()
    print(np.around(probs, 3))
```

# Evaluation

### Zero-Shot Classification
|  model   | dataset  | Top1 | Top5 |
|  ----  | ----  | ---- | ---- |
| TaiYi-CLIP-ViT-B-32-Roberta-Chinese  | ImageNet-CN | 40.64 % | 69.16% |

### Text-to-Image Retrieval

|  model   | dataset  | Top1 | Top5 | Top10 |
|  ----  | ----  | ---- | ---- | ---- |
| TaiYi-CLIP-ViT-B-32-Roberta-Chinese  | COCO-CN | 25.47 % | 51.70%  | 63.07% |
| TaiYi-CLIP-ViT-B-32-Roberta-Chinese  | wukong50k | 47.64 % | 80.97% | 89.51% |


# Citation

If you find the resource is useful, please cite the following website in your paper.

```
@misc{Fengshenbang-LM,
  title={Fengshenbang-LM},
  author={IDEA-CCNL},
  year={2022},
  howpublished={\url{https://github.com/IDEA-CCNL/Fengshenbang-LM}},
}
```