YOLOv10: Real-Time End-to-End Object Detection
Presents YOLOv10, an NMS-free, end-to-end real-time object detector with efficiency-accuracy driven design.
YOLOs balance speed and accuracy for real-time detection, but reliance on non-maximum suppression (NMS) blocks end-to-end deployment and raises latency, while component redundancy limits capability. YOLOv10 introduces consistent dual assignments for NMS-free training with competitive accuracy and low latency, plus a holistic efficiency-accuracy driven design of model components. It sets state-of-the-art results across scales; YOLOv10-S is 1.8x faster than RT-DETR-R18 at similar AP with 2.8x fewer parameters and FLOPs, and YOLOv10-B cuts latency 46% versus YOLOv9-C.
Based on: YOLOv10: Real-Time End-to-End Object Detection · Neural Information Processing Systems
Curated by Aramai Editorial
Read summary →