Index
YOLO v8
- YOLO #まとめ編
Architecture
概要
詳細
Backbone
YOLO v5 から利用されている CSPNet / Cross Stage Partial Network を利用.
- CSPNet / Cross Stage Partial Network
CSPNet の一部を C2f へ変更.
Neck
YOLO v5 と同様で、FPN (Feature Pyramid Network) + PAN (Path Aggregation Network)
Feature Pyramid Network
Path Aggregation Network
Head
Anchor Free / Decoupled Head に変更.
Anchor Free
Decoupled Head
精度
参考
開発元のUltralytics による論文はない.
Paper of YOLOv8
Real-Time Flying Object Detection with YOLOv8
- [2023]
- 技術レポート
- arxiv.org
What is YOLOv8: An In-Depth Exploration of the Internal Features of the Next-Generation Object Detector
- [2024]
- 3 Architectural Footprint of YOLOv8
- 3.6 CSPBackbone and Efficient Layer Aggregation
- arxiv.org
Web サイト
Ultralytics YOLOv8: State-of-the-Art YOLO Models
What is YOLOv8? A Complete Guide.
