Excited to announce RF-DETR, the current SOTA for real-time object detection, fully open source and Apache 2.0 for the community.
More to come but the repo and Colab notebook are available today for you to use
real-time AI use cases for wearables and mobile are now possible
we released RF-DETR nano, small and medium to give devs high accuracy models for low compute devices
nano is 11 mAP higher and .17 ms faster than recent YOLO releases
btw team @roboflow have been consistently beating the drum about how transformers have been eating vision for the last year+. if you're still on YOLO vWhatever, you're missing out.
check their appearance at last NeurIPS @latentspacepod live!
I can now detect made and missed shots in @NBA games
- detect jump-shots, layups and dunks
- classify as made or missed
- calculate distance from basket
- mark location on the court
big basketball release is coming soon to the roboflow/sports repo: github.com/roboflow/sports
over the weekend @chesterzelaya build cool drone detection demo; I decided to make "small improvements"
it's the same model
all I did was add supervision and trackers on top of it, two libraries I built
supervision: github.com/roboflow/super…
You can build AI systems like this using a fully open-source library, completely free. And it’s easier than you think.
Supervision simplifies computer vision workflows, making it easy to estimate vehicle speeds, track movement, and generate 3D visualizations. This data can drive
Over last weekend, my friends & I worked on building a bracket bot from scratch. We used 3D printed grippers with flexures to grip a box, an arm that extrude to press an elevator button. To recognize images of the elevator, Roboflow helps trains the bot to recognize the elevator.
Announcing Roboflow Universe: starting today, anyone can share their datasets and pre-trained models with the world.
We're supporting the community with new pricing which makes Roboflow's advanced features free to anyone publicly sharing their projects.