30/07/2021
Good news for developers and researchers!
Megvii has released YOLOX, an improved version of the acclaimed YOLO series, featuring a number of breakthroughs from the Megvii Research in the field of object . With improved performance and an engineering-friendly design, YOLOX aims to better bridge the gap between research and engineering applications.
In this version, we have optimized the YOLO detector to become anchor-free, while also integrating a number of advanced detection techniques, such as a decoupled head, strong data augmentation, and the leading label assignment strategy SimOTA to achieve state-of-the-art results.
We have also open source YOLOX and its deployment on MegEngine, ONNX, TensorRT, OpenVINO, and NCNN, to help global developers use and implement them more conveniently.
Megvii’s YOLO series aims to strike a balance between detection speed and accuracy. To find out more information and get started with your YOLOX implementation, visit the YOLOX GitHub at
(https://github.com/Megvii-BaseDetection/YOLOX).
YOLOX is a high-performance anchor-free YOLO, exceeding yolov3~v5 with MegEngine, ONNX, TensorRT, ncnn, and OpenVINO supported. - GitHub - Megvii-BaseDetection/YOLOX: YOLOX is a high-performance an...