深度学习论文阅读笔记(9)-R-CNN

Rich feature hierarchies for accurate object detection and semantic segmentation Tech report (v5)

Region proposals. "selective search"

feature extraction

Supervised pre-training

Domain-specific fine-tuning

We treat all region proposals with ≥ 0.5 IoU overlap with a ground-truth box as positives for that box’s class and the rest as negatives. (IOU 的阈值设的不太高，为了平衡正负样本，加大数据量)

SVM
<0.3 IoU 的是 负样本，0.3-0.7的忽略，

边框回归