Test_COCO_Captions

COCORetrieval

update

Pretrained CLIP models

The pt file trained by the baseline CLIP is formulated as

[visual backbone]_[trained dataset]_ep[epoch number].pt

The pt file trained by CLIP-KD is formulated as

[teacher visual backbone]_[teacher pretrained dataset]_kd_[student visual backbone]_[student trained dataset]_ep[epoch number].pt