File size: 14,084 Bytes
2d7eb6f fce6e7e a4cde0d 200cde7 fce6e7e a4cde0d 2d7eb6f f76d30f 2f14bfa f76d30f 1648844 f76d30f 1648844 f76d30f 1648844 f76d30f 1648844 f76d30f 1648844 f76d30f 1648844 f76d30f 1648844 f76d30f 1648844 f76d30f 1648844 f76d30f 1648844 f76d30f 1648844 f76d30f 1648844 f76d30f 1648844 f76d30f b357793 f76d30f 19c5e43 5d9c05f 19c5e43 5d9c05f 19c5e43 f76d30f e6525dd f76d30f |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 |
---
license: apache-2.0
widget:
- src: >-
https://huggingface.co/datasets/mishig/sample_images/resolve/main/cat-dog-music.png
candidate_labels: 音乐表演, 体育运动
example_title: 猫和狗
pipeline_tag: zero-shot-classification
---
[**中文说明**](README_CN.md) | [**English**](README.md)
# Introduction
This project aims to provide a better Chinese CLIP model. The training data used in this project consists of publicly accessible image URLs and related Chinese text descriptions, totaling 400 million. After screening, we ultimately used 100 million data for training.
This project is produced by QQ-ARC Joint Lab, Tencent PCG. We have also open-sourced our code on GitHub, [QA-CLIP](https://github.com/TencentARC-QQ/QA-CLIP), and welcome to star!
<br><br>
# Models and Results
<span id="model_card"></span>
## Model Card
QA-CLIP currently has three different open-source models of different sizes, and their model information and download links are shown in the table below:
<table border="1" width="100%">
<tr align="center">
<th>Model</th><th>Ckp</th><th>Params</th><th>Vision</th><th>Params of Vision</th><th>Text</th><th>Params of Text</th><th>Resolution</th>
</tr>
<tr align="center">
<td>QA-CLIP<sub>RN50</sub></td><td><a href="https://huggingface.co/TencentARC/QA-CLIP/resolve/main/QA-CLIP-RN50.pt">Download</a></td><td>77M</td><td>ResNet50</td><td>38M</td><td>RBT3</td><td>39M</td><td>224</td>
</tr>
<tr align="center">
<td>QA-CLIP<sub>ViT-B/16</sub></td><td><a href="https://huggingface.co/TencentARC/QA-CLIP/resolve/main/QA-CLIP-base.pt">Download</a></td><td>188M</td><td>ViT-B/16</td><td>86M</td><td>RoBERTa-wwm-Base</td><td>102M</td><td>224</td>
</tr>
<tr align="center">
<td>QA-CLIP<sub>ViT-L/14</sub></td><td><a href="https://huggingface.co/TencentARC/QA-CLIP/resolve/main/QA-CLIP-large.pt">Download</a></td><td>406M</td><td>ViT-L/14</td><td>304M</td><td>RoBERTa-wwm-Base</td><td>102M</td><td>224</td>
</tr>
</table>
<br>
## Results
We conducted zero-shot tests on [MUGE Retrieval](https://tianchi.aliyun.com/muge), [Flickr30K-CN](https://github.com/li-xirong/cross-lingual-cap), and [COCO-CN](https://github.com/li-xirong/coco-cn) datasets for image-text retrieval tasks. For the image zero-shot classification task, we tested on the ImageNet dataset. The test results are shown in the table below:
**Flickr30K-CN Zero-shot Retrieval (Official Test Set)**:
<table border="1" width="120%">
<tr align="center">
<th>Task</th><th colspan="3">Text-to-Image</th><th colspan="3">Image-to-Text</th>
</tr>
<tr align="center">
<td>Metric</td><td>R@1</td><td>R@5</td><td>R@10</td><td>R@1</td><td>R@5</td><td>R@10</td>
</tr>
<tr align="center">
<td width="120%">CN-CLIP<sub>RN50</sub></td><td>48.8</td><td>76.0</td><td>84.6</td><td>60.0</td><td>85.9</td><td>92.0</td>
</tr>
<tr align="center", style="background-color: Honeydew;">
<td width="120%">QA-CLIP<sub>RN50</sub></td><td><b>50.5</b></td><td><b>77.4</b></td><td><b>86.1</b></td><td><b>67.1</b></td><td><b>87.9</b></td><td><b>93.2</b></td>
</tr>
<tr align="center">
<td width="120%">CN-CLIP<sub>ViT-B/16</sub></td><td>62.7</td><td>86.9</td><td>92.8</td><td>74.6</td><td>93.5</td><td>97.1</td>
</tr>
<tr align="center", style="background-color: Honeydew;">
<td width="120%">QA-CLIP<sub>ViT-B/16</sub></td><td><b>63.8</b></td><td><b>88.0</b></td><td><b>93.2</b></td><td><b>78.4</b></td><td><b>96.1</b></td><td><b>98.5</b></td>
</tr>
<tr align="center">
<td width="120%">CN-CLIP<sub>ViT-L/14</sub></td><td>68.0</td><td>89.7</td><td>94.4</td><td>80.2</td><td>96.6</td><td>98.2</td>
</tr>
<tr align="center">
<td width="120%">AltClip<sub>ViT-L/14</sub></td><td><b>69.7</b></td><td>90.1</td><td><b>94.8</b></td><td>84.8</td><td>97.7</td><td>99.1</td>
</tr>
<tr align="center", style="background-color: Honeydew;">
<td width="120%">QA-CLIP<sub>ViT-L/14</sub></td><td>69.3</td><td><b>90.3</b></td><td>94.7</td><td><b>85.3</b></td><td><b>97.9</b></td><td><b>99.2</b></td>
</tr>
</table>
<br>
**MUGE Zero-shot Retrieval (Official Validation Set)**:
<table border="1" width="120%">
<tr align="center">
<th>Task</th><th colspan="3">Text-to-Image</th><th colspan="3">Image-to-Text</th>
</tr>
<tr align="center">
<td>Metric</td><td>R@1</td><td>R@5</td><td>R@10</td><td>R@1</td><td>R@5</td><td>R@10</td>
</tr>
<tr align="center">
<td width="120%">CN-CLIP<sub>RN50</sub></td><td>42.6</td><td>68.5</td><td>78.0</td><td>30.0</td><td>56.2</td><td>66.9</td>
</tr>
<tr align="center", style="background-color: Honeydew;">
<td width="120%">QA-CLIP<sub>RN50</sub></td><td><b>44.0</b></td><td><b>69.9</b></td><td><b>79.5</b></td><td><b>32.4</b></td><td><b>59.5</b></td><td><b>70.3</b></td>
</tr>
<tr align="center">
<td width="120%">CN-CLIP<sub>ViT-B/16</sub></td><td>52.1</td><td>76.7</td><td>84.4</td><td>38.7</td><td>65.6</td><td>75.1</td>
</tr>
<tr align="center", style="background-color: Honeydew;">
<td width="120%">QA-CLIP<sub>ViT-B/16</sub></td><td><b>53.2</b></td><td><b>77.7</b></td><td><b>85.1</b></td><td><b>40.7</b></td><td><b>68.2</b></td><td><b>77.2</b></td>
</tr>
<tr align="center">
<td width="120%">CN-CLIP<sub>ViT-L/14</sub></td><td>56.4</td><td>79.8</td><td>86.2</td><td>42.6</td><td>69.8</td><td>78.6</td>
</tr>
<tr align="center">
<td width="120%">AltClip<sub>ViT-L/14</sub></td><td>29.6</td><td>49.9</td><td>58.8</td><td>21.4</td><td>42.0</td><td>51.9</td>
</tr>
<tr align="center", style="background-color: Honeydew;">
<td width="120%">QA-CLIP<sub>ViT-L/14</sub></td><td><b>57.4</b></td><td><b>81.0</b></td><td><b>87.7</b></td><td><b>45.5</b></td><td><b>73.0</b></td><td><b>81.4</b></td>
</tr>
</table>
<br>
**COCO-CN Zero-shot Retrieval (Official Test Set)**:
<table border="1" width="120%">
<tr align="center">
<th>Task</th><th colspan="3">Text-to-Image</th><th colspan="3">Image-to-Text</th>
</tr>
<tr align="center">
<td>Metric</td><td>R@1</td><td>R@5</td><td>R@10</td><td>R@1</td><td>R@5</td><td>R@10</td>
</tr>
<tr align="center">
<td width="120%">CN-CLIP<sub>RN50</sub></td><td>48.1</td><td>81.3</td><td>90.5</td><td>50.9</td><td>81.1</td><td>90.5</td>
</tr>
<tr align="center", style="background-color: Honeydew;">
<td width="120%">QA-CLIP<sub>RN50</sub></td><td><b>50.1</b></td><td><b>82.5</b></td><td><b>91.7</b></td><td><b>56.7</b></td><td><b>85.2</b></td><td><b>92.9</b></td>
</tr>
<tr align="center">
<td width="120%">CN-CLIP<sub>ViT-B/16</sub></td><td>62.2</td><td>87.1</td><td>94.9</td><td>56.3</td><td>84.0</td><td>93.3</td>
</tr>
<tr align="center", style="background-color: Honeydew;">
<td width="120%">QA-CLIP<sub>ViT-B/16</sub></td><td><b>62.9</b></td><td><b>87.7</b></td><td><b>94.7</b></td><td><b>61.5</b></td><td><b>87.6</b></td><td><b>94.8</b></td>
</tr>
<tr align="center">
<td width="120%">CN-CLIP<sub>ViT-L/14</sub></td><td>64.9</td><td>88.8</td><td>94.2</td><td>60.6</td><td>84.4</td><td>93.1</td>
</tr>
<tr align="center">
<td width="120%">AltClip<sub>ViT-L/14</sub></td><td>63.5</td><td>87.6</td><td>93.5</td><td>62.6</td><td><b>88.5</b></td><td><b>95.9</b></td>
</tr>
<tr align="center", style="background-color: Honeydew;">
<td width="120%">QA-CLIP<sub>ViT-L/14</sub></td><td><b>65.7</b></td><td><b>90.2</b></td><td><b>95.0</b></td><td><b>64.5</b></td><td>88.3</td><td>95.1</td>
</tr>
</table>
<br>
**Zero-shot Image Classification on ImageNet**:
<table border="1" width="120%">
<tr align="center">
<th>Task</th><th colspan="1">ImageNet</th>
</tr>
<tr align="center">
<td width="120%">CN-CLIP<sub>RN50</sub></td><td>33.5</td>
</tr>
<tr align="center", style="background-color: Honeydew;">
<td width="120%">QA-CLIP<sub>RN50</sub></td><td><b>35.5</b></td>
</tr>
<tr align="center">
<td width="120%">CN-CLIP<sub>ViT-B/16</sub></td><td>48.4</td>
</tr>
<tr align="center", style="background-color: Honeydew;">
<td width="120%">QA-CLIP<sub>ViT-B/16</sub></td><td><b>49.7</b></td>
</tr>
<tr align="center">
<td width="120%">CN-CLIP<sub>ViT-L/14</sub></td><td>54.7</td>
</tr>
<tr align="center", style="background-color: Honeydew;">
<td width="120%">QA-CLIP<sub>ViT-L/14</sub></td><td><b>55.8</b></td>
</tr>
</table>
<br>
<br><br>
# Getting Started
## Installation Requirements
Environment configuration requirements:
* python >= 3.6.4
* pytorch >= 1.8.0 (with torchvision >= 0.9.0)
* CUDA Version >= 10.2
Install required packages:
```bash
cd /yourpath/QA-CLIP-main
pip install --upgrade pip
pip install -r requirements.txt
```
## Inference Code
```bash
export PYTHONPATH=/yourpath/QA-CLIP-main
```
Inference code example：
```python
import torch
from PIL import Image
import clip as clip
from clip import load_from_name, available_models
print("Available models:", available_models())
# Available models: ['ViT-B-16', 'ViT-L-14', 'RN50']
device = "cuda" if torch.cuda.is_available() else "cpu"
model, preprocess = load_from_name("ViT-B-16", device=device, download_root='./')
model.eval()
image = preprocess(Image.open("examples/pokemon.jpeg")).unsqueeze(0).to(device)
text = clip.tokenize(["杰尼龟", "妙蛙种子", "小火龙", "皮卡丘"]).to(device)
with torch.no_grad():
image_features = model.encode_image(image)
text_features = model.encode_text(text)
# Normalize the features. Please use the normalized features for downstream tasks.
image_features /= image_features.norm(dim=-1, keepdim=True)
text_features /= text_features.norm(dim=-1, keepdim=True)
logits_per_image, logits_per_text = model.get_similarity(image, text)
probs = logits_per_image.softmax(dim=-1).cpu().numpy()
print("Label probs:", probs)
```
<br><br>
## Prediction and Evaluation
### Download Image-text Retrieval Test Dataset
In Project <b>[Chinese-CLIP](https://github.com/OFA-Sys/Chinese-CLIP)</b>, the test set has already been preprocessed. Here is the download link they provided:
MUGE dataset：[download link](https://clip-cn-beijing.oss-cn-beijing.aliyuncs.com/datasets/MUGE.zip)
Flickr30K-CN dataset：[download link](https://clip-cn-beijing.oss-cn-beijing.aliyuncs.com/datasets/Flickr30k-CN.zip)
Additionally, obtaining the [COCO-CN](https://github.com/li-xirong/coco-cn) dataset requires applying to the original author.
### Download ImageNet Dataset
Please download the raw data yourself，[Chinese Label](http://clip-cn-beijing.oss-cn-beijing.aliyuncs.com/datasets/ImageNet-1K/label_cn.txt) and [English Label](http://clip-cn-beijing.oss-cn-beijing.aliyuncs.com/datasets/ImageNet-1K/label.txt) are provided by Project <b>[Chinese-CLIP](https://github.com/OFA-Sys/Chinese-CLIP)</b>
### Image-text Retrieval Evaluation
The image-text retrieval evaluation code can be referred to as follows:
```bash
split=test # Designate the computation of features for the valid or test set
resume=your_ckp_path
DATAPATH=your_DATAPATH
dataset_name=Flickr30k-CN
# dataset_name=MUGE
python -u eval/extract_features.py \
--extract-image-feats \
--extract-text-feats \
--image-data="${DATAPATH}/datasets/${dataset_name}/lmdb/${split}/imgs" \
--text-data="${DATAPATH}/datasets/${dataset_name}/${split}_texts.jsonl" \
--img-batch-size=32 \
--text-batch-size=32 \
--context-length=52 \
--resume=${resume} \
--vision-model=ViT-B-16 \
--text-model=RoBERTa-wwm-ext-base-chinese
python -u eval/make_topk_predictions.py \
--image-feats="${DATAPATH}/datasets/${dataset_name}/${split}_imgs.img_feat.jsonl" \
--text-feats="${DATAPATH}/datasets/${dataset_name}/${split}_texts.txt_feat.jsonl" \
--top-k=10 \
--eval-batch-size=32768 \
--output="${DATAPATH}/datasets/${dataset_name}/${split}_predictions.jsonl"
python -u eval/make_topk_predictions_tr.py \
--image-feats="${DATAPATH}/datasets/${dataset_name}/${split}_imgs.img_feat.jsonl" \
--text-feats="${DATAPATH}/datasets/${dataset_name}/${split}_texts.txt_feat.jsonl" \
--top-k=10 \
--eval-batch-size=32768 \
--output="${DATAPATH}/datasets/${dataset_name}/${split}_tr_predictions.jsonl"
python eval/evaluation.py \
${DATAPATH}/datasets/${dataset_name}/${split}_texts.jsonl \
${DATAPATH}/datasets/${dataset_name}/${split}_predictions.jsonl \
${DATAPATH}/datasets/${dataset_name}/output1.json
cat ${DATAPATH}/datasets/${dataset_name}/output1.json
python eval/transform_ir_annotation_to_tr.py \
--input ${DATAPATH}/datasets/${dataset_name}/${split}_texts.jsonl
python eval/evaluation_tr.py \
${DATAPATH}/datasets/${dataset_name}/${split}_texts.tr.jsonl \
${DATAPATH}/datasets/${dataset_name}/${split}_tr_predictions.jsonl \
${DATAPATH}/datasets/${dataset_name}/output2.json
cat ${DATAPATH}/datasets/${dataset_name}/output2.json
```
### ImageNet Zero-shot Classification
The ImageNet zero-shot classification code can be referred to as follows
```bash
bash scripts/zeroshot_eval.sh 0 \
${DATAPATH} imagenet \
ViT-B-16 RoBERTa-wwm-ext-base-chinese \
./pretrained_weights/QA-CLIP-base.pt
```
<br><br>
# Huggingface Model and Online Demo
We have open-sourced our model on the HuggingFace for easier access and utilization. Additionally, we have prepared a simple online demo for zero-shot classification, allowing everyone to experience it firsthand. We encourage you to give it a try!
[⭐️QA-CLIP-ViT-B-16⭐️](https://huggingface.co/TencentARC/QA-CLIP-ViT-B-16)
[⭐️QA-CLIP-ViT-L-14⭐️](https://huggingface.co/TencentARC/QA-CLIP-ViT-L-14)
Here are some examples for demonstration:
<br><br>
# Acknowledgments
The project code is based on implementation of <b>[Chinese-CLIP](https://github.com/OFA-Sys/Chinese-CLIP)</b>, and we are very grateful for their outstanding open-source contributions.
<br><br> |