Although the images were not labeled apart from a trigger word, you can have it generate higher quality versions of closeup photos of foods compared to the base model.
Trained on about 360 high quality images from the /r/foodporn subreddit
