WSL Embeddings

Facebook AI will be releasing embeddings from our weakly supervised learning models (WSL), allowing researchers and developers to experiment with image recognition capabilities based on high performance models that have been trained on a dataset of public images much larger than previously available. For example, one of the pre-trained computer vision models was trained using ResNext architecture with 101 layers on 3.5 billion images and 17,000 hashtags

Higher accuracy

Our WSL models have achieved a record-high score on ImageNet, a common benchmarking tool. Embeddings are planned to be provided for popular datasets like ImageNet, COCO, CUB2011, Oxford Flowers dataset, and more to help improve the performance of a diverse set of computer vision applications.

Get Started

Coming soon

More Tools


Detectron is a high-performance codebase for object detection, covering bounding box and object instance segmentation outputs.

Join Us

Tackle the world's most complex technology challenges

Join Our Team

Latest News

Visit the AI Blog for updates on recent publications, new tools, and more.

Visit Blog