Release of ImageNet-Captions
The json file contains a list of dictionaries describing the images in the dataset. Each image has the following information:
- filename: str
- wnid: str
- title: str
- description: str
- tags: list of str
Feel free to create your own captions, or just combine title, tag, description separated by spaces.
Paper link: http://arxiv.org/abs/2205.01397
NoCLIP Model link (35.7% ImageNet top-1): https://drive.google.com/file/d/100FHV8q75DLsef1l5_oJsdjIidyjp7RP/view?usp=sharing