Today's selection of cutting-edge AI information, welcome to read 👇
📖 Open-source OCR 2.0 model released, supporting multiple types of content recognition, with amazing results!
🕺 GVHMR technology accurately extracts human motions from videos, applicable in virtual reality and other fields!
🌟 awesome-LLM-resources compiles excellent resources on large language models, worth collecting and learning!
Cutting-edge Technology
1. An end-to-end open-source OCR model, called OCR 2.0!
It supports recognition of scene text, documents, notes, charts, mathematical formulas, and more, achieving a high BLEU score of 0.972.
GitHub: https://github.com/Ucas-HaoranWei/GOT-OCR2.0
Model download: https://huggingface.co/ucaslcl/GOT-OCR2_0
The model is only 1.43GB in size, and the demonstration results are incredibly impressive. Those interested can give it a try.
2. A technology that can precisely extract human motion information from videos: GVHMR.
For example, give GVHMR a video of a woman dancing, and it can analyze every movement and posture of the woman in the video, mapping them spatially and converting them into digital information for replication.
GitHub: https://github.com/zju3dv/GVHMR
Detailed introduction: https://zju3dv.github.io/gvhmr/
The application scenarios include virtual reality, film production, sports training, and even motion analysis in the medical field.
Open-source Projects
1. A collection of excellent resources related to large language models: awesome-LLM-resources.
It covers datasets, fine-tuning, inference, evaluation, RAG (retrieval-augmented generation), Agents, books, tutorials, papers, and more.
GitHub: https://github.com/WangRongsheng/awesome-LLM-resourses
The project is continuously being updated, aiming to gather the most comprehensive and up-to-date resources on large language models. It's worth following.
