v0.1.0 - First release!

• Standardized FID calculation: fixed image preprocessing and InceptionV3 model.
• FID-30k on MS-COCO validation set: we provide dataset on huggingface🤗, precomputed FID stats, fixed 30000 captions from MS-COCO that should be used to generate images
• Implementations of different popular text-to-image models to make metrics reproducible
• CLIP-score calculation
• User-friendly metrics calculation

v1.0.0-alpha

Benchmark for text2image models

First prerelease with main features:

• FID and CLIP score metrics
• Different dataloaders
• Models wrappers
• COCO dataset