Skip to content

Releases: mzhaoshuai/RLCF

Upload annotations, pre-triained weights of ClipCap and CapDec, etc

18 Jan 03:24
Compare
Choose a tag to compare
  • annotations.zip is the annotations you need for retrieval and captioning, including flickr30k, coco2014, nocaps.
  • flickr_train_set_image_text_vitb16_v2.zip is the CLIP-ViT-B/16 features of flickr30k. Extract it to get the .pkl file.
  • COCO_train_set_image_text_vitb16_v2.zip and COCO_train_set_image_text_vitl14.zip. CLIP features of coco2014.
  • clipcap_opt125m_transformer_coco_01.zip, pre-trained weights of ClipCap.
  • capdec_opt125m_transformer_coco_01.zip, pre-trained weights of CapDec.