r-cnn学习(六):RPN及AnchorTargetLayer学习

    RPN网络是faster与fast的主要区别,输入特征图,输出region proposals以及相应的分数。

  

# --------------------------------------------------------  
# Faster R-CNN  
# Copyright (c) 2015 Microsoft  
# Licensed under The MIT License [see LICENSE for details]  
# Written by Ross Girshick and Sean Bell  
# --------------------------------------------------------  
  
import os  
import caffe  
import yaml  
from fast_rcnn.config import cfg  
import numpy as np  
import numpy.random as npr  
from generate_anchors import generate_anchors  
from utils.cython_bbox import bbox_overlaps  
from fast_rcnn.bbox_transform import bbox_transform  
  
DEBUG = False  
  
class AnchorTargetLayer(caffe.Layer):  
    """ 
    Assign anchors to ground-truth targets. Produces anchor classification 
    labels and bounding-box regression targets. 
    """  
    #生成anchors,reshap输出
    def setup(self, bottom, top):  
        layer_params = yaml.load(self.param_str_)  
        anchor_scales = layer_params.get('scales', (8, 16, 32))  
        self._anchors = generate_anchors(scales=np.array(anchor_scales))#九个anchor的w h x_cstr y_cstr,对原始的wh做横向纵向变化,并放大缩小得到九个  
        self._num_anchors = self._anchors.shape[0]<span style="font-family: Arial, Helvetica, sans-serif;">#anchor的个数</span>  
        self._feat_stride = layer_params['feat_stride']#网络中参数16 (feature map为原图大小的1/16,1000*600->60*40) 
  
        if DEBUG:  
            print 'anchors:'  
            print self._anchors  
            print 'anchor shapes:'  
            print np.hstack((  
                self._anchors[:, 2::4] - self._anchors[:, 0::4],  
                self._anchors[:, 3::4] - self._anchors[:, 1::4],  
            ))  
            self._counts = cfg.EPS  
            self._sums = np.zeros((1, 4))  
            self._squared_sums = np.zeros((1, 4))  
            self._fg_sum = 0  
            self._bg_sum = 0  
            self._count = 0  
  
        # allow boxes to sit over the edge by a small amount  
        self._allowed_border = layer_params.get('allowed_border', 0)  
        #bottom 长度为4;bottom[0],map;bottom[1],boxes,labels;bottom[2],im_fo;bottom[3],图片数据  
        height, width = bottom[0].data.shape[-2:]  
        if DEBUG:  
            print 'AnchorTargetLayer: height', height, 'width', width  
  
        A = self._num_anchors#anchor的个数  
        # labels  
        top[0].reshape(1, 1, A * height, width)  
        # bbox_targets  
        top[1].reshape(1, A * 4, height, width)  
        # bbox_inside_weights  
        top[2].reshape(1, A * 4, height, width)  
        # bbox_outside_weights  
        top[3].reshape(1, A * 4, height, width)  
   #每个位置生成9个anchor,通过GT overlap过滤掉一部分anchors
    def forward(self, bottom, top):  
        # Algorithm:  
        #  
        # for each (H, W) location i  
        #   generate 9 anchor boxes centered on cell i  
        #   apply predicted bbox deltas at cell i to each of the 9 anchors  
        # filter out-of-image anchors  
        # measure GT overlap  
  
        assert bottom[0].data.shape[0] == 1, \  
            'Only single item batches are supported'  
  
#取得相应的anchors的h,w以及gt_box的位置,label # map of shape (..., H, W) height, width = bottom[0].data.shape[-2:] # GT boxes (x1, y1, x2, y2, label) gt_boxes = bottom[1].data#gt_boxes:长度不定 # im_info im_info = bottom[2].data[0, :] if DEBUG: print '' print 'im_size: ({}, {})'.format(im_info[0], im_info[1]) print 'scale: {}'.format(im_info[2]) print 'height, width: ({}, {})'.format(height, width) print 'rpn: gt_boxes.shape', gt_boxes.shape print 'rpn: gt_boxes', gt_boxes
#算出box的偏移量
# 1. Generate proposals from bbox deltas and shifted anchors shift_x = np.arange(0, width) * self._feat_stride shift_y = np.arange(0, height) * self._feat_stride shift_x, shift_y = np.meshgrid(shift_x, shift_y) shifts = np.vstack((shift_x.ravel(), shift_y.ravel(), shift_x.ravel(), shift_y.ravel())).transpose() # add A anchors (1, A, 4) to 根据偏移量移动anchors # cell K shifts (K, 1, 4) to get # shift anchors (K, A, 4) # reshape to (K*A, 4) shifted anchors A = self._num_anchors K = shifts.shape[0] all_anchors = (self._anchors.reshape((1, A, 4)) + shifts.reshape((1, K, 4)).transpose((1, 0, 2))) all_anchors = all_anchors.reshape((K * A, 4)) total_anchors = int(K * A)#K*A,所有anchors个数,包括越界的 #K: width*height #A: 9 # only keep anchors inside the image inds_inside = np.where( (all_anchors[:, 0] >= -self._allowed_border) & (all_anchors[:, 1] >= -self._allowed_border) & (all_anchors[:, 2] < im_info[1] + self._allowed_border) & # width (all_anchors[:, 3] < im_info[0] + self._allowed_border) # height )[0]#没有过界的anchors索引 if DEBUG: print 'total_anchors', total_anchors print 'inds_inside', len(inds_inside) # keep only inside anchors anchors = all_anchors[inds_inside, :]#没有过界的anchors if DEBUG: print 'anchors.shape', anchors.shape # label: 1 is positive, 0 is negative, -1 is dont care labels = np.empty((len(inds_inside), ), dtype=np.float32) labels.fill(-1) # overlaps between the anchors and the gt boxes # overlaps (ex, gt) overlaps = bbox_overlaps( #返回大小连续的overlaps,等同于排序 np.ascontiguousarray(anchors, dtype=np.float), np.ascontiguousarray(gt_boxes, dtype=np.float))
#找到某个box与所有gt_box最大的overlaps argmax_overlaps
= overlaps.argmax(axis=1)#overlaps每行最大值索引 max_overlaps = overlaps[np.arange(len(inds_inside)), argmax_overlaps]#最大的overlaps值
#找到某gt_box与所有box最大的overlaps gt_argmax_overlaps
= overlaps.argmax(axis=0) #overlaps每列中最大值索引 gt_max_overlaps = overlaps[gt_argmax_overlaps,#其对应的overlaps值 np.arange(overlaps.shape[1])] gt_argmax_overlaps = np.where(overlaps == gt_max_overlaps)[0] if not cfg.TRAIN.RPN_CLOBBER_POSITIVES: # assign bg labels first so that positive labels can clobber them labels[max_overlaps < cfg.TRAIN.RPN_NEGATIVE_OVERLAP] = 0 //对于某个gt,overlap最大的anchor为1 # fg label: for each gt, anchor with highest overlap labels[gt_argmax_overlaps] = 1 //对于某个anchor,其overlap超过某值为1 # fg label: above threshold IOU labels[max_overlaps >= cfg.TRAIN.RPN_POSITIVE_OVERLAP] = 1 if cfg.TRAIN.RPN_CLOBBER_POSITIVES: # assign bg labels last so that negative labels can clobber positives labels[max_overlaps < cfg.TRAIN.RPN_NEGATIVE_OVERLAP] = 0 # subsample positive labels if we have too many 如果正样本较多,降采样 num_fg = int(cfg.TRAIN.RPN_FG_FRACTION * cfg.TRAIN.RPN_BATCHSIZE) //正样本数量 fg_inds = np.where(labels == 1)[0] if len(fg_inds) > num_fg: disable_inds = npr.choice( fg_inds, size=(len(fg_inds) - num_fg), replace=False) labels[disable_inds] = -1 //多余正样本被随机标为负样本(这样真的好吗?) # subsample negative labels if we have too many 同样处理负样本 num_bg = cfg.TRAIN.RPN_BATCHSIZE - np.sum(labels == 1) bg_inds = np.where(labels == 0)[0] if len(bg_inds) > num_bg: disable_inds = npr.choice( bg_inds, size=(len(bg_inds) - num_bg), replace=False) labels[disable_inds] = -1 //仍然标为负? #print "was %s inds, disabling %s, now %s inds" % ( #len(bg_inds), len(disable_inds), np.sum(labels == 0)) #保留最大overlaps的anchors,其他为0(非极大值抑制?) bbox_targets = np.zeros((len(inds_inside), 4), dtype=np.float32) bbox_targets = _compute_targets(anchors, gt_boxes[argmax_overlaps, :]) ## #正样本inside_weights为1,其余为0(等同于论文中的pi* bbox_inside_weights = np.zeros((len(inds_inside), 4), dtype=np.float32) bbox_inside_weights[labels == 1, :] = np.array(cfg.TRAIN.RPN_BBOX_INSIDE_WEIGHTS) bbox_outside_weights = np.zeros((len(inds_inside), 4), dtype=np.float32)
#对样本权重进行归一化
if cfg.TRAIN.RPN_POSITIVE_WEIGHT < 0: # uniform weighting of examples (given non-uniform sampling) num_examples = np.sum(labels >= 0) positive_weights = np.ones((1, 4)) * 1.0 / num_examples negative_weights = np.ones((1, 4)) * 1.0 / num_examples else: assert ((cfg.TRAIN.RPN_POSITIVE_WEIGHT > 0) & (cfg.TRAIN.RPN_POSITIVE_WEIGHT < 1)) positive_weights = (cfg.TRAIN.RPN_POSITIVE_WEIGHT / np.sum(labels == 1)) negative_weights = ((1.0 - cfg.TRAIN.RPN_POSITIVE_WEIGHT) / np.sum(labels == 0)) bbox_outside_weights[labels == 1, :] = positive_weights bbox_outside_weights[labels == 0, :] = negative_weights if DEBUG: self._sums += bbox_targets[labels == 1, :].sum(axis=0) self._squared_sums += (bbox_targets[labels == 1, :] ** 2).sum(axis=0) self._counts += np.sum(labels == 1) means = self._sums / self._counts stds = np.sqrt(self._squared_sums / self._counts - means ** 2) print 'means:' print means print 'stdevs:' print stds # map up to original set of anchors 对total_anchors的其他box,weights及label进行填充 labels = _unmap(labels, total_anchors, inds_inside, fill=-1) bbox_targets = _unmap(bbox_targets, total_anchors, inds_inside, fill=0) bbox_inside_weights = _unmap(bbox_inside_weights, total_anchors, inds_inside, fill=0) bbox_outside_weights = _unmap(bbox_outside_weights, total_anchors, inds_inside, fill=0) if DEBUG: print 'rpn: max max_overlap', np.max(max_overlaps) print 'rpn: num_positive', np.sum(labels == 1) print 'rpn: num_negative', np.sum(labels == 0) self._fg_sum += np.sum(labels == 1) self._bg_sum += np.sum(labels == 0) self._count += 1 print 'rpn: num_positive avg', self._fg_sum / self._count print 'rpn: num_negative avg', self._bg_sum / self._count # labels 输出标签、box、inside_weights、outside_weights labels = labels.reshape((1, height, width, A)).transpose(0, 3, 1, 2) labels = labels.reshape((1, 1, A * height, width)) top[0].reshape(*labels.shape) top[0].data[...] = labels # bbox_targets bbox_targets = bbox_targets \ .reshape((1, height, width, A * 4)).transpose(0, 3, 1, 2) top[1].reshape(*bbox_targets.shape) top[1].data[...] = bbox_targets # bbox_inside_weights bbox_inside_weights = bbox_inside_weights \ .reshape((1, height, width, A * 4)).transpose(0, 3, 1, 2) assert bbox_inside_weights.shape[2] == height assert bbox_inside_weights.shape[3] == width top[2].reshape(*bbox_inside_weights.shape) top[2].data[...] = bbox_inside_weights # bbox_outside_weights bbox_outside_weights = bbox_outside_weights \ .reshape((1, height, width, A * 4)).transpose(0, 3, 1, 2) assert bbox_outside_weights.shape[2] == height assert bbox_outside_weights.shape[3] == width top[3].reshape(*bbox_outside_weights.shape) top[3].data[...] = bbox_outside_weights def backward(self, top, propagate_down, bottom): """This layer does not propagate gradients.""" pass def reshape(self, bottom, top): """Reshaping happens during the call to forward.""" pass def _unmap(data, count, inds, fill=0): """ Unmap a subset of item (data) back to the original set of items (of size count) """ #对于total_anchors,保留设定的label,其余填为fill if len(data.shape) == 1: ret = np.empty((count, ), dtype=np.float32) ret.fill(fill) ret[inds] = data else: ret = np.empty((count, ) + data.shape[1:], dtype=np.float32) ret.fill(fill) ret[inds, :] = data return ret def _compute_targets(ex_rois, gt_rois): """Compute bounding-box regression targets for an image.""" assert ex_rois.shape[0] == gt_rois.shape[0] assert ex_rois.shape[1] == 4 assert gt_rois.shape[1] == 5 return bbox_transform(ex_rois, gt_rois[:, :4]).astype(np.float32, copy=False)

 

  算偏移量时涉及到的公式:

     

这段代码主要生成anchors,算出anchors的偏移量,并根据与gt的overlaps,进行NMS及排序,赋予其相应的标签。

其中generate_anchors.py的源码如下。这段代码生成不同宽高比(1:2,1:1,2:1)、不同尺度(8 16 32)的anchors:

 

<span style="font-size:24px;">#功能描述:生成多尺度、多宽高比的anchors。  
#          尺度为:128,256,512; 宽高比为:1:2,1:1,2:1  
  
import numpy as np  #提供矩阵运算功能的库  
  
#生成anchors总函数:ratios为一个列表,表示宽高比为:1:2,1:1,2:1  
#2**x表示:2^x,scales:[2^3 2^4 2^5],即:[8 16 32]  
def generate_anchors(base_size=16, ratios=[0.5, 1, 2],  
                     scales=2**np.arange(3, 6)):  
    """ 
    Generate anchor (reference) windows by enumerating aspect ratios X 
    scales wrt a reference (0, 0, 15, 15) window. 
    """  
    base_anchor = np.array([1, 1, base_size, base_size]) - 1  #新建一个数组:base_anchor:[0 0 15 15]  
    ratio_anchors = _ratio_enum(base_anchor, ratios)  #枚举各种宽高比  
    anchors = np.vstack([_scale_enum(ratio_anchors[i, :], scales)  #枚举各种尺度,vstack:竖向合并数组  
                         for i in xrange(ratio_anchors.shape[0])]) #shape[0]:读取矩阵第一维长度,其值为3  
    return anchors  
  
#用于返回width,height,(x,y)中心坐标(对于一个anchor窗口)  
def _whctrs(anchor):  
    """ 
    Return width, height, x center, and y center for an anchor (window). 
    """  
    #anchor:存储了窗口左上角,右下角的坐标  
    w = anchor[2] - anchor[0] + 1  
    h = anchor[3] - anchor[1] + 1  
    x_ctr = anchor[0] + 0.5 * (w - 1)  #anchor中心点坐标  
    y_ctr = anchor[1] + 0.5 * (h - 1)  
    return w, h, x_ctr, y_ctr  
  
#给定一组宽高向量,输出各个anchor,即预测窗口,**输出anchor的面积相等,只是宽高比不同**  
def _mkanchors(ws, hs, x_ctr, y_ctr):  
    #ws:[23 16 11],hs:[12 16 22],ws和hs一一对应。  
    """ 
    Given a vector of widths (ws) and heights (hs) around a center 
    (x_ctr, y_ctr), output a set of anchors (windows). 
    """  
    ws = ws[:, np.newaxis]  #newaxis:将数组转置  
    hs = hs[:, np.newaxis]  
    anchors = np.hstack((x_ctr - 0.5 * (ws - 1),    #hstack、vstack:合并数组  
                         y_ctr - 0.5 * (hs - 1),    #anchor:[[-3.5 2 18.5 13]  
                         x_ctr + 0.5 * (ws - 1),     #        [0  0  15  15]  
                         y_ctr + 0.5 * (hs - 1)))     #       [2.5 -3 12.5 18]]  
    return anchors  
  
#枚举一个anchor的各种宽高比,anchor[0 0 15 15],ratios[0.5,1,2]  
def _ratio_enum(anchor, ratios):  
    """   列举关于一个anchor的三种宽高比 1:2,1:1,2:1 
    Enumerate a set of anchors for each aspect ratio wrt an anchor. 
    """  
  
    w, h, x_ctr, y_ctr = _whctrs(anchor)  #返回宽高和中心坐标,w:16,h:16,x_ctr:7.5,y_ctr:7.5  
    size = w * h   #size:16*16=256  
    size_ratios = size / ratios  #256/ratios[0.5,1,2]=[512,256,128]  
    #round()方法返回x的四舍五入的数字,sqrt()方法返回数字x的平方根  
    ws = np.round(np.sqrt(size_ratios)) #ws:[23 16 11]  
    hs = np.round(ws * ratios)    #hs:[12 16 22],ws和hs一一对应。as:23&12  
    anchors = _mkanchors(ws, hs, x_ctr, y_ctr)  #给定一组宽高向量,输出各个预测窗口  
    return anchors  
  
#枚举一个anchor的各种尺度,以anchor[0 0 15 15]为例,scales[8 16 32]  
def _scale_enum(anchor, scales):  
    """   列举关于一个anchor的三种尺度 128*128,256*256,512*512 
    Enumerate a set of anchors for each scale wrt an anchor. 
    """  
    w, h, x_ctr, y_ctr = _whctrs(anchor) #返回宽高和中心坐标,w:16,h:16,x_ctr:7.5,y_ctr:7.5  
    ws = w * scales   #[128 256 512]  
    hs = h * scales   #[128 256 512]  
    anchors = _mkanchors(ws, hs, x_ctr, y_ctr) #[[-56 -56 71 71] [-120 -120 135 135] [-248 -248 263 263]]  
    return anchors  
  
if __name__ == '__main__':  #主函数  
    import time  
    t = time.time()  
    a = generate_anchors()  #生成anchor(窗口)  
    print time.time() - t   #显示时间  
    print a  
    from IPython import embed; embed()  
</span>  

 

 

 

参考:http://blog.csdn.net/u010668907/article/details/51942481

          http://blog.csdn.net/xzzppp/article/details/52317863

posted @ 2016-12-07 15:53  牧马人夏峥  阅读(8104)  评论(1编辑  收藏  举报