BestHub
Sign in
Meituan Technology Team
Meituan Technology Team
Mar 2, 2023 · Artificial Intelligence

Technical Innovations in YOLOv6 3.0 for Real‑Time Object Detection

YOLOv6 3.0 introduces RepBi‑PAN neck, Anchor‑Aided Training, and Decoupled Location Distillation, achieving 57.2% AP at 29 FPS while improving small‑object detection with under 4% speed loss, and the paper provides extensive ablations and practical guidance for researchers and engineers developing high‑performance real‑time object detectors.

YOLOv6anchor trainingcomputer vision
0 likes · 12 min read
Technical Innovations in YOLOv6 3.0 for Real‑Time Object Detection
Meituan Technology Team
Meituan Technology Team
Sep 22, 2022 · Artificial Intelligence

Quantization Deployment Scheme for YOLOv6: Methods, Optimizations, and Performance Evaluation

The paper proposes a full quantization pipeline for YOLOv6 that combines a re‑parameterization optimizer, partial PTQ, channel‑wise distillation, graph‑scale merging, and GPU‑offloaded preprocessing, enabling an INT8 model to retain ~42 % mAP while delivering over 200 % throughput increase and 40 % QPS gain versus FP16.

Channel DistillationPTQPerformance optimization
0 likes · 16 min read
Quantization Deployment Scheme for YOLOv6: Methods, Optimizations, and Performance Evaluation
Meituan Technology Team
Meituan Technology Team
Sep 15, 2022 · Artificial Intelligence

YOLOv6 2.0: Enhanced Object Detection Models and Quantization Solutions

YOLOv6 2.0, released on September 5 2022, adds a CSPStackRep backbone, self‑distillation and a RepOptimizer‑based quantization pipeline that let the lightweight YOLOv6‑S achieve 869 FPS with 43.3 mAP, while the M and L models reach up to 52.5 % COCO AP at 121–233 FPS, halve training time, and support end‑to‑end deployment on TensorRT, OpenVINO and ARM devices.

AIYOLOv6computer vision
0 likes · 5 min read
YOLOv6 2.0: Enhanced Object Detection Models and Quantization Solutions
Meituan Technology Team
Meituan Technology Team
Jun 23, 2022 · Artificial Intelligence

YOLOv6: An Efficient Industrial Object Detection Framework

YOLOv6, an open‑source industrial object detection framework from Meituan Visual Intelligence, combines a hardware‑friendly EfficientRep backbone, Rep‑PAN neck, and Efficient Decoupled Head with anchor‑free training, SimOTA assignment, and SIoU loss, delivering COCO AP up to 43.1% at over 500 FPS and supporting TensorRT, OpenVINO, MNN, TNN, and NCNN deployment.

Efficient InferenceYOLOv6anchor-free
0 likes · 13 min read
YOLOv6: An Efficient Industrial Object Detection Framework