9. proposal_target_layer_tf.py ( Faster-RCNN_TF代码解读)

9. /lib/rpn_msr/proposal_target_layer_tf.py

本文件实现了根据max_overlaps筛选前景和背景，并得到bbox regression targets等。
具体思路如下（右边页是思路，左边页是备注）：

# --------------------------------------------------------
# Faster R-CNN
# Copyright (c) 2015 Microsoft
# Licensed under The MIT License [see LICENSE for details]
# Written by Ross Girshick and Sean Bell
# --------------------------------------------------------

import yaml
import numpy as np
import numpy.random as npr
from fast_rcnn.config import cfg
from fast_rcnn.bbox_transform import bbox_transform
from utils.cython_bbox import bbox_overlaps
import pdb

DEBUG = False
#传入的数据为
# rpn_rois：blob，内容为[proposal引索(全零)，proposal]
#gt_boxes：gtound-truth
#_num_classes：类别总数，21
#函数作用：
# 产生筛选后的roi，对应labels，三个（len(rois),4*21）大小的矩阵，其中一个对fg-roi对应引索行的对应类别的4个位置填上（dx,dy,dw,dh），另两个对fg-roi对应引索行的对应类别的4个位置填上（1,1,1,1）
def proposal_target_layer(rpn_rois, gt_boxes,_num_classes):
    """
    Assign object detection proposals to ground-truth targets. Produces proposal
    classification labels and bounding-box regression targets.
    """
    # Proposal ROIs (0, x1, y1, x2, y2) coming from RPN
    # (i.e., rpn.proposal_layer.ProposalLayer), or any other source
    all_rois = rpn_rois
    # TODO(rbg): it's annoying that sometimes I have extra info before
    # and other times after box coordinates -- normalize to one format

    # Include ground-truth boxes in the set of candidate rois
    #建立一个（gt_boxes.shape[0]+proposals.shape[0],5）的array，proposals信息在上GT信息在下面，存入all_rois，GT引索全初始化为0
    zeros = np.zeros((gt_boxes.shape[0], 1), dtype=gt_boxes.dtype)