File size: 654 Bytes
92e0882 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 |
# Alpha-CLIP evaluation
## Zero-Shot Classification on ImageNet-S
1.prepare [imagenet-s](https://github.com/LUSSeg/ImageNet-S) dataset, only `validation` raw image is needed.
2.download [imagenet_919.json](https://download.openxlab.org.cn/models/SunzeY/AlphaCLIP/weight/imagenet_919.json) we provide as data annotation (generated from imagenet-s annotation). The folder should be structured like
```
├── imagenet_s_zs_test
│ ├── data
│ │ ├── imagenet_919.json
│ │ └── ImageNetS919
│ │ └── validation
```
3.run test script.
```
cd eval/imagenet_s_zs_test
python imagenet_s_zs_test.py
```
|