Closed rB080 closed 9 months ago
Hi, in our extended version "Building an open-vocabulary video clip model with better architectures, optimization and data", we have constructed captions for kinetics videos with the help of BLIP-2 and LLaMA-2 chat for improved performance on zero-shot action recognition and video-text retrieval.
The k400 caption file have been uploaded to the 'blip_llama2_caption' folder at present, and the subsequent code will be updated soon.
Thank you so much for the captions file!
Is there a csv/json/txt/ etc file containing captions for all kinetics videos?