Suggestions
MB J.
Research Assistant @ University of Applied Sciences Dortmund | Research and Development (R&D)
I am the founder of allSTEAM, a social enterprise that provides free access to high-quality and locally adapted STEAM educational resources to students and learners in developing countries. With over four years of experience in leading and managing digital transformation and product growth projects, I help organizations leverage the power of AI, IoT, and edge computing ethically and responsibly.
I am passionate about the application of artificial intelligence in healthcare, especially for low-resource languages and communities. I co-organize tinyML, a global initiative that promotes the development and deployment of ultra-low power machine learning solutions. I also mentor and coach young African leaders to develop entrepreneurial and productive mindsets and skills. I constantly seek and share knowledge with my community, and I am dedicated to creating the largest and most skilled workforce in Africa.
Highlights
Meta Reality Labs present Sapiens: /...
"Sapiens: Foundation for Human Vision Models" is super cool paper with some seriously impressive results for the following tasks:-
- 2D pose estimation ( Improved mAP by 7.6 on Humans-5K)
- Body-part segmentation (Improved mIoU by 17.1 on Humans-2K)
- Depth estimation : (Reduced RMSE by 22.4% on Hi4D)
- Surface Normal Estimation: (Reduced angular error by 53.5% on THuman2)
Just think about it, training on 300 million human images and then fine-tuning to nail all these tasks with state-of-the-art performance. That’s wild, right?
Real-world applications
- VR/AR.
- In healthcare and sport:- physical therapy, fitness tracking, body movement for rehabilitation.
- For security and surveillance task
- In robotics to help improve human gesture and movement.
- In gaming; media production; and etc.
If you're into computer vision, especially anything human-centric, you’ve gotta check this out. The potential here for real-world applications is just huge.
You can checkout the GitHub repo here: https://lnkd.in/ewx2Swbn
Qwen team has released its latest vision language model - the SOTA open-source Qwen2-VL series. I tried the demo, and I must say, its impressive. you can try it too on Hugging Face https://lnkd.in/e9FiCnjT
#LargeLanguageModels #Qwen #VisionLanguageModels #Innovation
