| # 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 | |
| ``` | |