The map now includes speech, robotics, and document reading.
Alongside Language-specific Datasets, we added 13 categories since our last newsletter.
Speech & Audio includes OpenAI's Whisper, which turns speech into text and translates spoken languages into English. Robotics & Embodied AI includes Hugging Face's LeRobot, which lets people record demonstrations and use them to train robots. Document Conversion & OCR includes Docling, a tool developed at IBM Research that extracts text and tables from documents such as PDFs.
These additions let you examine open and closed options for more tasks, from transcribing a recording to reading a scanned document. The map explains what each product does and what its creators allow others to use, modify, and share. Explore the new categories.