RedCaps: web-curated image-text data created by the people, for the people
This provides a high-quality, scalable data source for vision and vision-language tasks, though it is incremental in leveraging social media for data collection.
The authors tackled the problem of noisy web data for image-text pairs by introducing RedCaps, a 12M dataset collected from Reddit, which enabled training models that produced human-preferred captions and transferred well to downstream tasks.
Large datasets of paired images and text have become increasingly popular for learning generic representations for vision and vision-and-language tasks. Such datasets have been built by querying search engines or collecting HTML alt-text -- since web data is noisy, they require complex filtering pipelines to maintain quality. We explore alternate data sources to collect high quality data with minimal filtering. We introduce RedCaps -- a large-scale dataset of 12M image-text pairs collected from Reddit. Images and captions from Reddit depict and describe a wide variety of objects and scenes. We collect data from a manually curated set of subreddits, which give coarse image labels and allow us to steer the dataset composition without labeling individual instances. We show that captioning models trained on RedCaps produce rich and varied captions preferred by humans, and learn visual representations that transfer to many downstream tasks.