datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
nla-bottleneck-resultsDescribe-then-Generate-BottleneckThe Describe-then-Generate-Bottleneck dataset investigates the phenomenon of information loss when AI systems generate images from textual descriptions of existing images. This dataset contains 150 randomly selected samples that demonstrate the "bottleneck" effect occurring in describe-then-generate pipelines.
The dataset explores a critical question in AI vision: How much visual information is lost when we ask a vision-language model to describe an image, and then use that description to… See the full description on the dataset page: https://huggingface.co/datasets/sportsvision/Describe-then-Generate-Bottleneck.nla-bottleneck-domains