ML Tasks Image Detection YOLO V3

[junxnone]

YOLO V3

Reference

• 2015 YOLO V1 You Only Look Once: Unified, Real-Time Object Detection [paper] [netscope] [Code] [Code2 - AlexeyAB]
• 2016 YOLO V2 YOLO9000: Better, Faster, Stronger [paper] [netscope] [Code]
• 2018 YOLOv3 An Incremental Improvement [paper] [YOLOv3 netscope] [Code]
• implements
  • 官方实现 - darknet yolo
  • fork 后 update 较多的code - AlexeyAB
  • Minimal PyTorch implementation of YOLOv3
  • keras-YOLOv3-mobilenet
  • keras-yolo3
  • kmeans-anchor-boxes
  • keras-yolo3-detection
• 目标检测（九）--YOLO v1,v2,v3
• YOLO lite
• YOLO-LITE项目实现(比SSD和MobileNet更快的算法)
• pjreddie Publications
• YOLO-v3模型参数anchor设置
• YOLOv2 YOLOv3 如何选择先验框（priors anchor）（自用）
• YOLO v3网络结构分析
• YOLOV3 基于OpenCV DNN 的目标检测实现
• 目标检测之YOLO V2 V3
• 从yolov1到yolov3
• YOLO V3
• Tiny Darknet

Brief

• darknet53
  • 引入残差结构
• 引入 FPN - 多尺度检测
  • 52x52
  • 26x26
  • 13x13
• multilabel classification : independent logistic 替换 softmax -> binary cross-entropy loss

---

其他无效tricks

• Anchor box x, y offset predictions
• Linear x, y predictions instead of logistic
• Focal Loss
• Dual IOU thresholds and truth assignment

darknet53

FPN

COCO dataset 9 clusters

• (10×13)
• (16×30)
• (33×23)
• (30×61)
• (62×45)
• (59×119)
• (116 × 90)
• (156 × 198)
• (373 × 326)

[junxnone]

junxnone/aiwiki#53