YOLO官網:
YOLO v.s Faster R-CNN:
1.統一網絡:YOLO沒有顯示求取region proposal的過程。Faster R-CNN中儘管RPN與fast rcnn共享卷積層,但是在模型訓練過程中,需要反覆訓練RPN網絡和fast rcnn網絡.相對於R-CNN系列的"看兩眼"(候選框提取與分類),YOLO只需要Look Once.
2. YOLO統一爲一個迴歸問題,而R-CNN將檢測結果分爲兩部分求解:物體類別(分類問題),物體位置即bounding box(迴歸問題)。
2. YOLOv1: You Only Look Once: Unified, Real-Time Object Detection
目標檢測之YOLO v1算法: You Only Look Once: Unified, Real-Time Object Detection:
3. YOLOv2 (YOLO9000: Better, Faster, Stronger)
目標檢測之YOLOv2 算法-YOLO9000: Better, Faster, Stronger:
4. YOLOv3: An Incremental Improvement
目標檢測之YOLOv3算法: An Incremental Improvement:
5. Tiny YOLOv3
目標檢測之Tiny YOLOv3算法:
6. YOLOv4: Optimal Speed and Accuracy of Object Detection
目標檢測之YOLOv4算法: Optimal Speed and Accuracy of Object Detection:
7. YOLOv5算法
目標檢測之YOLOv5算法:
8. YOLObile算法
YOLObile:面向移動設備的「實時目標檢測」算法(AAAI 2021):
9. YOLOF算法
YOLOF:You Only Look One-level Feature(CVPR 2021):
10. YOLOX算法
YOLOX: Exceeding YOLO Series in 2021
增添目標檢測數據集PASCAL VOC和COCO詳細解析:
- 目標檢測數據集PASCAL VOC詳解:
2. 目標檢測數據集MSCOCO詳解:
參考
- ^V1,V2,V3參考地址: https://blog.csdn.net/App_12062011/article/details/77554288
- ^V4轉載地址: https://mp.weixin.qq.com/s/Ua3T-DOuzmLWuXfohEiVFw
- ^一文讀懂YOLO V5 與 YOLO V4 https://zhuanlan.zhihu.com/p/161083602
- ^近距離觀察YOLOv3 https://zhuanlan.zhihu.com/p/40332004
- ^Faster-RCNN的anchor和YOLOv3的anchor一樣嗎 https://blog.csdn.net/xiqi4145/article/details/86516511