faster RCNN(keras版本)代码讲解博客索引:
1.faster RCNN(keras版本)代码讲解(1)-概述
2.faster RCNN(keras版本)代码讲解(2)-数据准备
3.faster RCNN(keras版本)代码讲解(3)-训练流程详情
4.faster RCNN(keras版本)代码讲解(4)-共享卷积层详情
5.faster RCNN(keras版本)代码讲解(5)-RPN层详情
6.faster RCNN(keras版本)代码讲解(6)-ROI Pooling层详情
一.RPN层网络
其实就是ResNet的卷积层+一层Conv,然后进行分类(该框是否是物体),回归(定位一个框的四个坐标值)
def rpn(base_layers,num_anchors):
x = Convolution2D(512, (3, 3), padding='same', activation='relu', kernel_initializer='normal', name='rpn_conv1')(base_layers)
#对于rpn的2分类使用的是sigmoid激活函数,所以这里的filter数量为num_anchors,而论文中使用的softmax,所以论文中filter数量为num_anchors*2,使用1*1卷积代替全连接
x_class = Convolution2D(num_anchors, (1, 1), activation='sigmoid', kernel_initializer='uniform', name='rpn_out_class')(x)
x_regr = Convolution2D(num_anchors * 4, (1, 1), activation='linear', kernel_initializer='zero', name='rpn_out_regress')(x)
return [x_class, x_regr, base_layers]
二.RPN网络层的标签
1.根据特征图的大小,对每个特征图上的点,反推回原始图像,然后在原始图像上找到框
2.如果该框和实际的框gt IOU>=0.7,则认为该框为正例,如果IOU<=0.3,则认为为背景,如果0.3大于IOU小于0.7则将该框丢弃
# RPN,生成region proposal
def calc_rpn(C, img_data, width, height, resized_width, resized_height, img_length_calc_function):
#downscale = 16,因为共享卷积层会将resize后的图片压缩16倍
downscale = float(C.rpn_stride)
anchor_sizes = C.anchor_box_scales #[128, 256, 512]
anchor_ratios = C.anchor_box_ratios
num_anchors = len(anchor_sizes) * len(anchor_ratios)
# calculate the output map size based on the network architecture
# 计算图片经过共享卷积以后输出的特征图的大小
(output_width, output_height) = img_length_calc_function(resized_width, resized_height)
print("经过共享卷积层后特征图大小after share conv",output_height,output_width)
n_anchratios = len(anchor_ratios)
# initialise empty output objectives
#保存和gt 是否重叠(是正例还是反例),根据IOU值,output_height, output_width中的内容对应在其中一种模板框下框的坐标,总体表示为在这个坐标下,是否和gt重叠,重叠为1正例,不重叠为0反例,num_anchors表示厚度,表示9重不同规模的框
y_rpn_overlap = np.zeros((output_height, output_width, num_anchors))
#保存 框是否启用
y_is_box_valid = np.zeros((output_height, output_width, num_anchors))
#保存 框的4个坐标值
y_rpn_regr = np.zeros((output_height, output_width, num_anchors * 4))
num_bboxes = len(img_data['bboxes'])
print("图片中物体个数",num_bboxes)
num_anchors_for_bbox = np.zeros(num_bboxes).astype(int)
best_anchor_for_bbox = -1*np.ones((num_bboxes, 4)).astype(int)
best_iou_for_bbox = np.zeros(num_bboxes).astype(np.float32)
best_x_for_bbox = np.zeros((num_bboxes, 4)).astype(int)
best_dx_for_bbox = np.zeros((num_bboxes, 4)).astype(np.float32)
# get the GT box coordinates, and resize to account for image resizing
gta = np.zeros((num_bboxes, 4))
for bbox_num, bbox in enumerate(img_data['bboxes']):
# get the GT box coordinates, and resize to account for image resizing
#重新获得resize过后物体的位置
gta[bbox_num, 0] = bbox['x1'] * (resized_width / float(width))
gta[bbox_num, 1] = bbox['x2'] * (resized_width / float(width))
gta[bbox_num, 2] = bbox['y1'] * (resized_height / float(height))
gta[bbox_num, 3] = bbox['y2'] * (resized_height / float(height))
# rpn ground truth
# 对9种框进行遍历,里面还有个对图片中所有物体的遍历,所有for循环次数 len(anchor_sizes)
# 遍历顺序,先遍历框,然后拿特征图和原始图框坐标,然后再和原始图中物体进行IOU计算,拿IOU最高的框
for anchor_size_idx in range(len(anchor_sizes)):
for anchor_ratio_idx in range(n_anchratios):
anchor_x = anchor_sizes[anchor_size_idx] * anchor_ratios[anchor_ratio_idx][0]
anchor_y = anchor_sizes[anchor_size_idx] * anchor_ratios[anchor_ratio_idx][1]
# 根据特征图的大小,对每个特征图上的点,反推回原始图像,然后在原始图像上找到框
for ix in range(output_width):
# x-coordinates of the current anchor box
x1_anc = downscale * (ix + 0.5) - anchor_x / 2
x2_anc = downscale * (ix + 0.5) + anchor_x / 2
# ignore boxes that go across image boundaries
# 超过边界的框就不要了
if x1_anc < 0 or x2_anc > resized_width:
continue
for jy in range(output_height):
# y-coordinates of the current anchor box
y1_anc = downscale * (jy + 0.5) - anchor_y / 2
y2_anc = downscale * (jy + 0.5) + anchor_y / 2
# ignore boxes that go across image boundaries
if y1_anc < 0 or y2_anc > resized_height:
continue
# bbox_type indicates whether an anchor should be a target
bbox_type = 'neg' #negative
# this is the best IOU for the (x,y) coord and the current anchor
# note that this is different from the best IOU for a GT bbox
best_iou_for_loc = 0.0
for bbox_num in range(num_bboxes):
# get IOU of the current GT box and the current anchor box
# 对框进行一个IOU计算,IOU越高,越说明候选框anchor和gt重叠比列越高,越接近gt
curr_iou = iou([gta[bbox_num, 0], gta[bbox_num, 2], gta[bbox_num, 1], gta[bbox_num, 3]], [x1_anc, y1_anc, x2_anc, y2_anc])
# calculate the regression targets if they will be needed
# 这里没看懂,求gt和框的中心点重叠程度?
if curr_iou > best_iou_for_bbox[bbox_num] or curr_iou > C.rpn_max_overlap:
cx = (gta[bbox_num, 0] + gta[bbox_num, 1]) / 2.0
cy = (gta[bbox_num, 2] + gta[bbox_num, 3]) / 2.0
cxa = (x1_anc + x2_anc)/2.0
cya = (y1_anc + y2_anc)/2.0
tx = (cx - cxa) / (x2_anc - x1_anc)
ty = (cy - cya) / (y2_anc - y1_anc)
tw = np.log((gta[bbox_num, 1] - gta[bbox_num, 0]) / (x2_anc - x1_anc))
th = np.log((gta[bbox_num, 3] - gta[bbox_num, 2]) / (y2_anc - y1_anc))
if img_data['bboxes'][bbox_num]['class'] != 'bg':
# all GT boxes should be mapped to an anchor box, so we keep track of which anchor box was best
if curr_iou > best_iou_for_bbox[bbox_num]:
#保存当前最好的特征图上的anchor坐标,注意是当前
best_anchor_for_bbox[bbox_num] = [jy, ix, anchor_ratio_idx, anchor_size_idx]
#保存gt对应当前最好anchor的IOU值
best_iou_for_bbox[bbox_num] = curr_iou
#保存当前最好anchor对应在原图上的坐标
best_x_for_bbox[bbox_num,:] = [x1_anc, x2_anc, y1_anc, y2_anc]
#这个应该是留着反向传播用吧
best_dx_for_bbox[bbox_num,:] = [tx, ty, tw, th]
# we set the anchor to positive if the IOU is >0.7 (it does not matter if there was another better box, it just indicates overlap)
# 如果IOU>0.7,我们就设置为这个anchor为正例
if curr_iou > C.rpn_max_overlap:
bbox_type = 'pos' #positive
#记录每个物体框有几个anchor有较高重叠率IOU>0.7的
num_anchors_for_bbox[bbox_num] += 1
# we update the regression layer target if this IOU is the best for the current (x,y) and anchor position
if curr_iou > best_iou_for_loc:
best_iou_for_loc = curr_iou
best_regr = (tx, ty, tw, th)
# if the IOU is >0.3 and <0.7, it is ambiguous and no included in the objective
#if IOU is >0.3 and <0.7,就丢弃这些anchor
if C.rpn_min_overlap < curr_iou < C.rpn_max_overlap:
# gray zone between neg and pos
if bbox_type != 'pos':
bbox_type = 'neutral'
# turn on or off outputs depending on IOUs
if bbox_type == 'neg':
y_is_box_valid[jy, ix, anchor_ratio_idx + n_anchratios * anchor_size_idx] = 1
y_rpn_overlap[jy, ix, anchor_ratio_idx + n_anchratios * anchor_size_idx] = 0
elif bbox_type == 'neutral':
y_is_box_valid[jy, ix, anchor_ratio_idx + n_anchratios * anchor_size_idx] = 0
y_rpn_overlap[jy, ix, anchor_ratio_idx + n_anchratios * anchor_size_idx] = 0
elif bbox_type == 'pos':
y_is_box_valid[jy, ix, anchor_ratio_idx + n_anchratios * anchor_size_idx] = 1
y_rpn_overlap[jy, ix, anchor_ratio_idx + n_anchratios * anchor_size_idx] = 1
start = 4 * (anchor_ratio_idx + n_anchratios * anchor_size_idx)
y_rpn_regr[jy, ix, start:start+4] = best_regr
# we ensure that every bbox has at least one positive RPN region
# 确保每一个物体框都有一个正例
for idx in range(num_anchors_for_bbox.shape[0]):
if num_anchors_for_bbox[idx] == 0:
# no box with an IOU greater than zero ...
if best_anchor_for_bbox[idx, 0] == -1:
continue
y_is_box_valid[
best_anchor_for_bbox[idx,0], best_anchor_for_bbox[idx,1], best_anchor_for_bbox[idx,2] + n_anchratios *
best_anchor_for_bbox[idx,3]] = 1
y_rpn_overlap[
best_anchor_for_bbox[idx,0], best_anchor_for_bbox[idx,1], best_anchor_for_bbox[idx,2] + n_anchratios *
best_anchor_for_bbox[idx,3]] = 1
start = 4 * (best_anchor_for_bbox[idx,2] + n_anchratios * best_anchor_for_bbox[idx,3])
y_rpn_regr[
best_anchor_for_bbox[idx,0], best_anchor_for_bbox[idx,1], start:start+4] = best_dx_for_bbox[idx, :]
y_rpn_overlap = np.transpose(y_rpn_overlap, (2, 0, 1))
y_rpn_overlap = np.expand_dims(y_rpn_overlap, axis=0)
y_is_box_valid = np.transpose(y_is_box_valid, (2, 0, 1))
y_is_box_valid = np.expand_dims(y_is_box_valid, axis=0)
y_rpn_regr = np.transpose(y_rpn_regr, (2, 0, 1))
y_rpn_regr = np.expand_dims(y_rpn_regr, axis=0)
pos_locs = np.where(np.logical_and(y_rpn_overlap[0, :, :, :] == 1, y_is_box_valid[0, :, :, :] == 1))
neg_locs = np.where(np.logical_and(y_rpn_overlap[0, :, :, :] == 0, y_is_box_valid[0, :, :, :] == 1))
num_pos = len(pos_locs[0])
# one issue is that the RPN has many more negative than positive regions, so we turn off some of the negative
# regions. We also limit it to 256 regions.
#因为RPN中negative要比positive regions多,所以要剪掉一些negative,然后现在regions的总量256,正例不能超过128
num_regions = 256
if len(pos_locs[0]) > num_regions/2:
val_locs = random.sample(range(len(pos_locs[0])), len(pos_locs[0]) - num_regions/2)
y_is_box_valid[0, pos_locs[0][val_locs], pos_locs[1][val_locs], pos_locs[2][val_locs]] = 0
num_pos = num_regions/2
if len(neg_locs[0]) + num_pos > num_regions:
val_locs = random.sample(range(len(neg_locs[0])), len(neg_locs[0]) - num_pos)
y_is_box_valid[0, neg_locs[0][val_locs], neg_locs[1][val_locs], neg_locs[2][val_locs]] = 0
y_rpn_cls = np.concatenate([y_is_box_valid, y_rpn_overlap], axis=1)
y_rpn_regr = np.concatenate([np.repeat(y_rpn_overlap, 4, axis=1), y_rpn_regr], axis=1)
print("特征图上每点(其实是点映射的框)对应的类别y_rpn_cls",y_rpn_cls.shape)
print("特征图上每点(其实是点映射的框)对应的4个坐标y_rpn_regr",y_rpn_regr.shape)
return np.copy(y_rpn_cls), np.copy(y_rpn_regr)
三.RPN层损失函数的计算
#RPN分类损失,先传递框的数量
def rpn_loss_cls(num_anchors):
def rpn_loss_cls_fixed_num(y_true, y_pred):
if K.image_dim_ordering() == 'tf':
#K.sum(y_true[:, :, :, :num_anchors]为框是否启用,binary_crossentropy采用交叉熵,/K.sum(epsilon + y_true[:, :, :, :num_anchors]),再求个平均
return lambda_rpn_class * K.sum(y_true[:, :, :, :num_anchors] * K.binary_crossentropy(y_pred[:, :, :, :], y_true[:, :, :, num_anchors:])) / K.sum(epsilon + y_true[:, :, :, :num_anchors])
else:
return lambda_rpn_class * K.sum(y_true[:, :num_anchors, :, :] * K.binary_crossentropy(y_pred[:, :, :, :], y_true[:, num_anchors:, :, :])) / K.sum(epsilon + y_true[:, :num_anchors, :, :])
return rpn_loss_cls_fixed_num