Faster-RCNN模型结构解析

最新推荐文章于 2023-01-19 21:04:10 发布
最新推荐文章于 2023-01-19 21:04:10 发布
阅读量2.9k 收藏 9
点赞数
分类专栏： 深度学习
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/bblingbbling/article/details/106919687
版权

Faster-RCNN模型结构解析

一、网络结构

（论文地址： https://arxiv.org/pdf/1506.01497.pdf）
1.backbone：VGG、Resnet、Xception等
2.neck：无
3.head：sparse-prediction
图1
在这里插入图片描述

二、backbone

FasterRcnn对不限制输入图片的尺寸，但是一般会把输入进来的图片短边固定成600，另一边等比例resize。

Resnet50特征提取网络

代码实现

def ResNet50(inputs):

    img_input = inputs

    x = ZeroPadding2D((3, 3))(img_input)
    x = Conv2D(64, (7, 7), strides=(2, 2), name='conv1')(x)
    x = BatchNormalization(name='bn_conv1')(x)
    x = Activation('relu')(x)

    x = MaxPooling2D((3, 3), strides=(2, 2), padding="same")(x)

    x = conv_block(x, 3, [64, 64, 256], stage=2, block='a', strides=(1, 1))
    x = identity_block(x, 3, [64, 64, 256], stage=2, block='b')
    x = identity_block(x, 3, [64, 64, 256], stage=2, block='c')


    x = conv_block(x, 3, [128, 128, 512], stage=3, block='a')
    x = identity_block(x, 3, [128, 128, 512], stage=3, block='b')
    x = identity_block(x, 3, [128, 128, 512], stage=3, block='c')
    x = identity_block(x, 3, [128, 128, 512], stage=3, block='d')

    x = conv_block(x, 3, [256, 256, 1024], stage=4, block='a')
    x = identity_block(x, 3, [256, 256, 1024], stage=4, block='b')
    x = identity_block(x, 3, [256, 256, 1024], stage=4, block='c')
    x = identity_block(x, 3, [256, 256, 1024], stage=4, block='d')
    x = identity_block(x, 3, [256, 256, 1024], stage=4, block='e')
    x = identity_block(x, 3, [256, 256, 1024], stage=4, block='f')

    return x

参数计算

input_shape = (600, 600, 3)        output_shape = (38, 38, 1024)
__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
input_1 (InputLayer)            (None, 600, 600, 3)  0                                            
__________________________________________________________________________________________________
zero_padding2d_1 (ZeroPadding2D (None, 606, 606, 3)  0           input_1[0][0]                    
__________________________________________________________________________________________________
conv1 (Conv2D)                  (None, 300, 300, 64) 9472        zero_padding2d_1[0][0]           
__________________________________________________________________________________________________
bn_conv1 (BatchNormalization)   (None, 300, 300, 64) 256         conv1[0][0]                      
__________________________________________________________________________________________________
activation_1 (Activation)       (None, 300, 300, 64) 0           bn_conv1[0][0]                   
__________________________________________________________________________________________________
max_pooling2d_1 (MaxPooling2D)  (None, 150, 150, 64) 0           activation_1[0][0]               
__________________________________________________________________________________________________
res2a_branch2a (Conv2D)         (None, 150, 150, 64) 4160        max_pooling2d_1[0][0]            
__________________________________________________________________________________________________
bn2a_branch2a (BatchNormalizati (None, 150, 150, 64) 256         res2a_branch2a[0][0]             
__________________________________________________________________________________________________
activation_2 (Activation)       (None, 150, 150, 64) 0           bn2a_branch2a[0][0]              
__________________________________________________________________________________________________
res2a_branch2b (Conv2D)         (None, 150, 150, 64) 36928       activation_2[0][0]               
__________________________________________________________________________________________________
bn2a_branch2b (BatchNormalizati (None, 150, 150, 64) 256         res2a_branch2b[0][0]             
__________________________________________________________________________________________________
activation_3 (Activation)       (None, 150, 150, 64) 0           bn2a_branch2b[0][0]              
__________________________________________________________________________________________________
res2a_branch2c (Conv2D)         (None, 150, 150, 256 16640       activation_3[0][0]               
__________________________________________________________________________________________________
res2a_branch1 (Conv2D)          (None, 150, 150, 256 16640       max_pooling2d_1[0][0]            
__________________________________________________________________________________________________
bn2a_branch2c (BatchNormalizati (None, 150, 150, 256 1024        res2a_branch2c[0][0]             
__________________________________________________________________________________________________
bn2a_branch1 (BatchNormalizatio (None, 150, 150, 256 1024        res2a_branch1[0][0]              
__________________________________________________________________________________________________
add_1 (Add)                     (None, 150, 150, 256 0           bn2a_branch2c[0][0]              
                                                                 bn2a_branch1[0][0]               
__________________________________________________________________________________________________
activation_4 (Activation)       (None, 150, 150, 256 0           add_1[0][0]                      
__________________________________________________________________________________________________
res2b_branch2a (Conv2D)         (None, 150, 150, 64) 16448       activation_4[0][0]               
__________________________________________________________________________________________________
bn2b_branch2a (BatchNormalizati (None, 150, 150, 64) 256         res2b_branch2a[0][0]             
__________________________________________________________________________________________________
activation_5 (Activation)       (None, 150, 150, 64) 0           bn2b_branch2a[0][0]              
__________________________________________________________________________________________________
res2b_branch2b (Conv2D)         (None, 150, 150, 64) 36928       activation_5[0][0]               
__________________________________________________________________________________________________
bn2b_branch2b (BatchNormalizati (None, 150, 150, 64) 256         res2b_branch2b[0][0]             
__________________________________________________________________________________________________
activation_6 (Activation)       (None, 150, 150, 64) 0           bn2b_branch2b[0][0]              
__________________________________________________________________________________________________
res2b_branch2c (Conv2D)         (None, 150, 150, 256 16640       activation_6[0][0]               
__________________________________________________________________________________________________
bn2b_branch2c (BatchNormalizati (None, 150, 150, 256 1024        res2b_branch2c[0][0]             
__________________________________________________________________________________________________
add_2 (Add)                     (None, 150, 150, 256 0           bn2b_branch2c[0][0]              
                                                                 activation_4[0][0]               
__________________________________________________________________________________________________
activation_7 (Activation)       (None, 150, 150, 256 0           add_2[0][0]                      
__________________________________________________________________________________________________
res2c_branch2a (Conv2D)         (None, 150, 150, 64) 16448       activation_7[0][0]               
__________________________________________________________________________________________________
bn2c_branch2a (BatchNormalizati (None, 150, 150, 64) 256         res2c_branch2a[0][0]             
__________________________________________________________________________________________________
activation_8 (Activation)       (None, 150, 150, 64) 0           bn2c_branch2a[0][0]              
__________________________________________________________________________________________________
res2c_branch2b (Conv2D)         (None, 150, 150, 64) 36928       activation_8[0][0]               
__________________________________________________________________________________________________
bn2c_branch2b (BatchNormalizati (None, 150, 150, 64) 256         res2c_branch2b[0][0]             
__________________________________________________________________________________________________
activation_9 (Activation)       (None, 150, 150, 64) 0           bn2c_branch2b[0][0]              
__________________________________________________________________________________________________
res2c_branch2c (Conv2D)         (None, 150, 150, 256 16640       activation_9[0][0]               
__________________________________________________________________________________________________
bn2c_branch2c (BatchNormalizati (None, 150, 150, 256 1024        res2c_branch2c[0][0]             
__________________________________________________________________________________________________
add_3 (Add)                     (None, 150, 150, 256 0           bn2c_branch2c[0][0]              
                                                                 activation_7[0][0]               
__________________________________________________________________________________________________
activation_10 (Activation)      (None, 150, 150, 256 0           add_3[0][0]                      
__________________________________________________________________________________________________
res3a_branch2a (Conv2D)         (None, 75, 75, 128)  32896       activation_10[0][0]              
__________________________________________________________________________________________________
bn3a_branch2a (BatchNormalizati (None, 75, 75, 128)  512         res3a_branch2a[0][0]             
__________________________________________________________________________________________________
activation_11 (Activation)      (None, 75, 75, 128)  0           bn3a_branch2a[0][0]              
__________________________________________________________________________________________________
res3a_branch2b (Conv2D)         (None, 75, 75, 128)  147584      activation_11[0][0]              
__________________________________________________________________________________________________
bn3a_branch2b (BatchNormalizati (None, 75, 75, 128)  512         res3a_branch2b[0][0]             
__________________________________________________________________________________________________
activation_12 (Activation)      (None, 75, 75, 128)  0           bn3a_branch2b[0][0]              
__________________________________________________________________________________________________
res3a_branch2c (Conv2D)         (None, 75, 75, 512)  66048       activation_12[0][0]              
__________________________________________________________________________________________________
res3a_branch1 (Conv2D)          (None, 75, 75, 512)  131584      activation_10[0][0]              
__________________________________________________________________________________________________
bn3a_branch2c (BatchNormalizati (None, 75, 75, 512)  2048        res3a_branch2c[0][0]             
__________________________________________________________________________________________________
bn3a_branch1 (BatchNormalizatio (None, 75, 75, 512)  2048        res3a_branch1[0][0]              
__________________________________________________________________________________________________
add_4 (Add)                     (None, 75, 75, 512)  0           bn3a_branch2c[0][0]              
                                                                 bn3a_branch1[0][0]               
__________________________________________________________________________________________________
activation_13 (Activation)      (None, 75, 75, 512)  0           add_4[0][0]                      
__________________________________________________________________________________________________
res3b_branch2a (Conv2D)         (None, 75, 75, 128)  65664       activation_13[0][0]              
__________________________________________________________________________________________________
bn3b_branch2a (BatchNormalizati (None, 75, 75, 128)  512         res3b_branch2a[0][0]             
__________________________________________________________________________________________________
activation_14 (Activation)      (None, 75, 75, 128)  0           bn3b_branch2a[0][0]              
__________________________________________________________________________________________________
res3b_branch2b (Conv2D)         (None, 75, 75, 128)  147584      activation_14[0][0]              
__________________________________________________________________________________________________
bn3b_branch2b (BatchNormalizati (None, 75, 75, 128)  512         res3b_branch2b[0][0]             
__________________________________________________________________________________________________
activation_15 (Activation)      (None, 75, 75, 128)  0           bn3b_branch2b[0][0]              
__________________________________________________________________________________________________
res3b_branch2c (Conv2D)         (None, 75, 75, 512)  66048       activation_15[0][0]              
__________________________________________________________________________________________________
bn3b_branch2c (BatchNormalizati (None, 75, 75, 512)  2048        res3b_branch2c[0][0]             
__________________________________________________________________________________________________
add_5 (Add)                     (None, 75, 75, 512)  0           bn3b_branch2c[0][0]              
                                                                 activation_13[0][0]              
__________________________________________________________________________________________________
activation_16 (Activation)      (None, 75, 75, 512)  0           add_5[0][0]                      
__________________________________________________________________________________________________
res3c_branch2a (Conv2D)         (None, 75, 75, 128)  65664       activation_16[0][0]              
__________________________________________________________________________________________________
bn3c_branch2a (BatchNormalizati (None, 75, 75, 128)  512         res3c_branch2a[0][0]             
__________________________________________________________________________________________________
activation_17 (Activation)      (None, 75, 75, 128)  0           bn3c_branch2a[0][0]              
__________________________________________________________________________________________________
res3c_branch2b (Conv2D)         (None, 75, 75, 128)  147584      activation_17[0][0]              
__________________________________________________________________________________________________
bn3c_branch2b (BatchNormalizati (None, 75, 75, 128)  512         res3c_branch2b[0][0]             
__________________________________________________________________________________________________
activation_18 (Activation)      (None, 75, 75, 128)  0           bn3c_branch2b[0][0]              
__________________________________________________________________________________________________
res3c_branch2c (Conv2D)         (None, 75, 75, 512)  66048       activation_18[0][0]              
__________________________________________________________________________________________________
bn3c_branch2c (BatchNormalizati (None, 75, 75, 512)  2048        res3c_branch2c[0][0]             
__________________________________________________________________________________________________
add_6 (Add)                     (None, 75, 75, 512)  0           bn3c_branch2c[0][0]              
                                                                 activation_16[0][0]              
__________________________________________________________________________________________________
activation_19 (Activation)      (None, 75, 75, 512)  0           add_6[0][0]                      
__________________________________________________________________________________________________
res3d_branch2a (Conv2D)         (None, 75, 75, 128)  65664       activation_19[0][0]              
__________________________________________________________________________________________________
bn3d_branch2a (BatchNormalizati (None, 75, 75, 128)  512         res3d_branch2a[0][0]             
__________________________________________________________________________________________________
activation_20 (Activation)      (None, 75, 75, 128)  0           bn3d_branch2a[0][0]              
__________________________________________________________________________________________________
res3d_branch2b (Conv2D)         (None, 75, 75, 128)  147584      activation_20[0][0]              
__________________________________________________________________________________________________
bn3d_branch2b (BatchNormalizati (None, 75, 75, 128)  512         res3d_branch2b[0][0]             
__________________________________________________________________________________________________
activation_21 (Activation)      (None, 75, 75, 128)  0           bn3d_branch2b[0][0]              
__________________________________________________________________________________________________
res3d_branch2c (Conv2D)         (None, 75, 75, 512)  66048       activation_21[0][0]              
__________________________________________________________________________________________________
bn3d_branch2c (BatchNormalizati (None, 75, 75, 512)  2048        res3d_branch2c[0][0]             
__________________________________________________________________________________________________
add_7 (Add)                     (None, 75, 75, 512)  0           bn3d_branch2c[0][0]              
                                                                 activation_19[0][0]              
__________________________________________________________________________________________________
activation_22 (Activation)      (None, 75, 75, 512)  0           add_7[0][0]                      
__________________________________________________________________________________________________
res4a_branch2a (Conv2D)         (None, 38, 38, 256)  131328      activation_22[0][0]              
__________________________________________________________________________________________________
bn4a_branch2a (BatchNormalizati (None, 38, 38, 256)  1024        res4a_branch2a[0][0]             
__________________________________________________________________________________________________
activation_23 (Activation)      (None, 38, 38, 256)  0           bn4a_branch2a[0][0]              
__________________________________________________________________________________________________
res4a_branch2b (Conv2D)         (None, 38, 38, 256)  590080      activation_23[0][0]              
__________________________________________________________________________________________________
bn4a_branch2b (BatchNormalizati (None, 38, 38, 256)  1024        res4a_branch2b[0][0]             
__________________________________________________________________________________________________
activation_24 (Activation)      (None, 38, 38, 256)  0           bn4a_branch2b[0][0]              
__________________________________________________________________________________________________
res4a_branch2c (Conv2D)         (None, 38, 38, 1024) 263168      activation_24[0][0]              
__________________________________________________________________________________________________
res4a_branch1 (Conv2D)          (None, 38, 38, 1024) 525312      activation_22[0][0]              
__________________________________________________________________________________________________
bn4a_branch2c (BatchNormalizati (None, 38, 38, 1024) 4096        res4a_branch2c[0][0]             
__________________________________________________________________________________________________
bn4a_branch1 (BatchNormalizatio (None, 38, 38, 1024) 4096        res4a_branch1[0][0]              
__________________________________________________________________________________________________
add_8 (Add)                     (None, 38, 38, 1024) 0           bn4a_branch2c[0][0]              
                                                                 bn4a_branch1[0][0]               
__________________________________________________________________________________________________
activation_25 (Activation)      (None, 38, 38, 1024) 0           add_8[0][0]                      
__________________________________________________________________________________________________
res4b_branch2a (Conv2D)         (None, 38, 38, 256)  262400      activation_25[0][0]              
__________________________________________________________________________________________________
bn4b_branch2a (BatchNormalizati (None, 38, 38, 256)  1024        res4b_branch2a[0][0]             
__________________________________________________________________________________________________
activation_26 (Activation)      (None, 38, 38, 256)  0           bn4b_branch2a[0][0]              
__________________________________________________________________________________________________
res4b_branch2b (Conv2D)         (None, 38, 38, 256)  590080      activation_26[0][0]              
__________________________________________________________________________________________________
bn4b_branch2b (BatchNormalizati (None, 38, 38, 256)  1024        res4b_branch2b[0][0]             
__________________________________________________________________________________________________
activation_27 (Activation)      (None, 38, 38, 256)  0           bn4b_branch2b[0][0]              
__________________________________________________________________________________________________
res4b_branch2c (Conv2D)         (None, 38, 38, 1024) 263168      activation_27[0][0]              
__________________________________________________________________________________________________
bn4b_branch2c (BatchNormalizati (None, 38, 38, 1024) 4096        res4b_branch2c[0][0]             
__________________________________________________________________________________________________
add_9 (Add)                     (None, 38, 38, 1024) 0           bn4b_branch2c[0][0]              
                                                                 activation_25[0][0]              
__________________________________________________________________________________________________
activation_28 (Activation)      (None, 38, 38, 1024) 0           add_9[0][0]                      
__________________________________________________________________________________________________
res4c_branch2a (Conv2D)         (None, 38, 38, 256)  262400      activation_28[0][0]              
__________________________________________________________________________________________________
bn4c_branch2a (BatchNormalizati (None, 38, 38, 256)  1024        res4c_branch2a[0][0]             
__________________________________________________________________________________________________
activation_29 (Activation)      (None, 38, 38, 256)  0           bn4c_branch2a[0][0]              
__________________________________________________________________________________________________
res4c_branch2b (Conv2D)         (None, 38, 38, 256)  590080      activation_29[0][0]              
__________________________________________________________________________________________________
bn4c_branch2b (BatchNormalizati (None, 38, 38, 256)  1024        res4c_branch2b[0][0]             
__________________________________________________________________________________________________
activation_30 (Activation)      (None, 38, 38, 256)  0           bn4c_branch2b[0][0]              
__________________________________________________________________________________________________
res4c_branch2c (Conv2D)         (None, 38, 38, 1024) 263168      activation_30[0][0]              
__________________________________________________________________________________________________
bn4c_branch2c (BatchNormalizati (None, 38, 38, 1024) 4096        res4c_branch2c[0][0]             
__________________________________________________________________________________________________
add_10 (Add)                    (None, 38, 38, 1024) 0           bn4c_branch2c[0][0]              
                                                                 activation_28[0][0]              
__________________________________________________________________________________________________
activation_31 (Activation)      (None, 38, 38, 1024) 0           add_10[0][0]                     
__________________________________________________________________________________________________
res4d_branch2a (Conv2D)         (None, 38, 38, 256)  262400      activation_31[0][0]              
__________________________________________________________________________________________________
bn4d_branch2a (BatchNormalizati (None, 38, 38, 256)  1024        res4d_branch2a[0][0]             
__________________________________________________________________________________________________
activation_32 (Activation)      (None, 38, 38, 256)  0           bn4d_branch2a[0][0]              
__________________________________________________________________________________________________
res4d_branch2b (Conv2D)         (None, 38, 38, 256)  590080      activation_32[0][0]              
__________________________________________________________________________________________________
bn4d_branch2b (BatchNormalizati (None, 38, 38, 256)  1024        res4d_branch2b[0][0]             
__________________________________________________________________________________________________
activation_33 (Activation)      (None, 38, 38, 256)  0           bn4d_branch2b[0][0]              
__________________________________________________________________________________________________
res4d_branch2c (Conv2D)         (None, 38, 38, 1024) 263168      activation_33[0][0]              
__________________________________________________________________________________________________
bn4d_branch2c (BatchNormalizati (None, 38, 38, 1024) 4096        res4d_branch2c[0][0]             
__________________________________________________________________________________________________
add_11 (Add)                    (None, 38, 38, 1024) 0           bn4d_branch2c[0][0]              
                                                                 activation_31[0][0]              
__________________________________________________________________________________________________
activation_34 (Activation)      (None, 38, 38, 1024) 0           add_11[0][0]                     
__________________________________________________________________________________________________
res4e_branch2a (Conv2D)         (None, 38, 38, 256)  262400      activation_34[0][0]              
__________________________________________________________________________________________________
bn4e_branch2a (BatchNormalizati (None, 38, 38, 256)  1024        res4e_branch2a[0][0]             
__________________________________________________________________________________________________
activation_35 (Activation)      (None, 38, 38, 256)  0           bn4e_branch2a[0][0]              
__________________________________________________________________________________________________
res4e_branch2b (Conv2D)         (None, 38, 38, 256)  590080      activation_35[0][0]              
__________________________________________________________________________________________________
bn4e_branch2b (BatchNormalizati (None, 38, 38, 256)  1024        res4e_branch2b[0][0]             
__________________________________________________________________________________________________
activation_36 (Activation)      (None, 38, 38, 256)  0           bn4e_branch2b[0][0]              
__________________________________________________________________________________________________
res4e_branch2c (Conv2D)         (None, 38, 38, 1024) 263168      activation_36[0][0]              
__________________________________________________________________________________________________
bn4e_branch2c (BatchNormalizati (None, 38, 38, 1024) 4096        res4e_branch2c[0][0]             
__________________________________________________________________________________________________
add_12 (Add)                    (None, 38, 38, 1024) 0           bn4e_branch2c[0][0]              
                                                                 activation_34[0][0]              
__________________________________________________________________________________________________
activation_37 (Activation)      (None, 38, 38, 1024) 0           add_12[0][0]                     
__________________________________________________________________________________________________
res4f_branch2a (Conv2D)         (None, 38, 38, 256)  262400      activation_37[0][0]              
__________________________________________________________________________________________________
bn4f_branch2a (BatchNormalizati (None, 38, 38, 256)  1024        res4f_branch2a[0][0]             
__________________________________________________________________________________________________
activation_38 (Activation)      (None, 38, 38, 256)  0           bn4f_branch2a[0][0]              
__________________________________________________________________________________________________
res4f_branch2b (Conv2D)         (None, 38, 38, 256)  590080      activation_38[0][0]              
__________________________________________________________________________________________________
bn4f_branch2b (BatchNormalizati (None, 38, 38, 256)  1024        res4f_branch2b[0][0]             
__________________________________________________________________________________________________
activation_39 (Activation)      (None, 38, 38, 256)  0           bn4f_branch2b[0][0]              
__________________________________________________________________________________________________
res4f_branch2c (Conv2D)         (None, 38, 38, 1024) 263168      activation_39[0][0]              
__________________________________________________________________________________________________
bn4f_branch2c (BatchNormalizati (None, 38, 38, 1024) 4096        res4f_branch2c[0][0]             
__________________________________________________________________________________________________
add_13 (Add)                    (None, 38, 38, 1024) 0           bn4f_branch2c[0][0]              
                                                                 activation_37[0][0]              
__________________________________________________________________________________________________
activation_40 (Activation)      (None, 38, 38, 1024) 0           add_13[0][0]                     
==================================================================================================
Total params: 8,589,184
Trainable params: 8,528,000
Non-trainable params: 61,184
__________________________________________________________________________________________________

三、RPN

600x600的图片经过特征提取获取共享特征层后，feature maps的先验框数量为38x38x9=12996个，这些先验框会经过两路卷积，分别是9x2的分类卷积(判断获得的先验框是否包含物体)和9x4的回归卷积(对这些先验框进行调整，获得一个新box)，目的是为了获取包含obj的建议框，即对图片中哪块区域有物体进行初步筛选。
在这里插入图片描述

def get_rpn(base_layers, num_anchors):
    x = Conv2D(512, (3, 3), padding='same', activation='relu', kernel_initializer='normal', name='rpn_conv1')(base_layers)

    x_class = Conv2D(num_anchors, (1, 1), activation='sigmoid', kernel_initializer='uniform', name='rpn_out_class')(x)
    x_regr = Conv2D(num_anchors * 4, (1, 1), activation='linear', kernel_initializer='zero', name='rpn_out_regress')(x)
    
    x_class = Reshape((-1,1),name="classification")(x_class)
    x_regr = Reshape((-1,4),name="regression")(x_regr)
    return [x_class, x_regr, base_layers]

三、RoiPooling

作用就是将一个个不同大小的region proposals从feature maps上抠出来，进行池化，输出到固定大小的矩形框。
1.对38x38x1024的fearure maps的建议框按数量每次次默认为32输入roipooling；
2.将每批次的建议框resize至(32,14,14,1024)的尺寸；
———————————————————————
3.通过一次Conv Block stride=2压缩至(32,7,7,2048)；
4.通过两次Identity Block加深至(32,7,7,2048)；
5.通过AveragePooling2D kernel=7压缩至(32,1,1,2048)；
6.Flatten至(32,2048)；
7.Dense(32, num_classes)和Dense(32, (num_classes-1)x4).

四、Loss

loss = rpn_cls + rpn_reg + roi_cls + roi_reg

其中：
rpn_cls判断有无目标
rpn_reg对anchor box进行微调
roi_cls判断目标类别
roi_reg对目标box位置精调

reg损失函数使用了soomth L1 Loss(0点附近更加平滑)；cls损失函数使用了cross entropy。

bblingbbling
关注
  • 0
    点赞
  • 踩
  • 9
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
专栏目录
FasterRCNN模型原理与实现
禅与计算机程序设计艺术
04-03 949
FasterR-CNN模型原理与实现 作者：禅与计算机程序设计艺术 1. 背景介绍 物体检测是计算机视觉领域的一项重要任务,它旨在准确地识别和定位图像中的目标对象。其中,深度学习方法在物体检测领域取得了突破性进展,其中R-CNN、Fast R-CNN和Faster R-C
faster rcnn 网络结构图解
10-15
faster rcnn 网络结构图解
Faster-RCNN网络
最白の白菜的博客
03-02 3208
Faster-RCNN网络 学习目标 熟悉FasterRCNN目标检测的思想 知道anchor（锚框）的思想 掌握RPN网络是如何进行候选区域的生成的 掌握ROIPooling的使用方法 知道fasterRCNN的训练方法 在R-CNN和Fast RCNN的基础上，在2016年提出了Faster RCNN网络模型，在结构上，Faster RCNN已经将候选区域的生成，特征提取，目标分类及目标框的回归都整合在了一个网络中，综合性能有较大提高，在检测速度方面尤为明显。接下来我们给大家详细介绍fasterR
Faster RCNN 结构总结
hugegrunt的博客
05-06 3013
本文重点介绍Faster RCNN的结构以及其中的一些注意事项。重点以偏重直观理解为主，不求严谨。对于Faster RCNN网络的训练方法并未在本文中提及。Fast RCNN提出了将SS找出的ROI映射到CNN网络提取出的feature map上的方法。而Faster RCNN 也继承了这一思路，并且在此基础上对ROI的提取方式做了改进。
Faster-rcnn详解
热门推荐
WZZ18191171661的博客
03-04 3万+
论文题目：Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks 论文链接：论文链接 论文代码：Matlab版本点击此处，Python版本点击此处 作为一个目标检测领域的baseline算法，Faster-rcnn值得你去仔细理解里面的细节，如果你能深入的了解这些，我相信你会受益匪浅。...
FasterRCNN之整体框架详解
littlehaes的博客
10-24 1万+
欢迎访问我的个人主页 训练时迭代了50000次,pascal2007测试结果MAP=0.65,其中,chair的ap最低0.46,horse的ap最高0.81 运行环境:CUDA8,cuDNN7,1070Ti,TensorFlow1.4.0,python2.7 训练集:voc_2007_trainval 训练时间148分钟 测试集:voc_2007_testt 测试时间10分钟(5000图) 刚接...
faster rcnn结构及代码讲解
m0_37663944的博客
12-26 1808
本篇博客主要讲解faster rcnn的基本结构及相关代码讲解。 总体结构介绍   从上图中我们可以看出，faster rcnn一共有三个部分，我们大致先说下：   1.第一个部分为特征提取部分，经过卷积层得到特征图，也就是feature map   2.第二个部分为RPN(region proposal network)区域候选网络，这是对fast rcnn重点改进的一部分，它的主要作用是得到...
Faster-RCNN-with-torchvision:使用Torchvision重新实现Faster-rcnn
05-04
代码博客解析地址：模型地址：安装代码及环境搭建$ git clone https://github.com/supernotman/Faster-RCNN-with-torchvision.git$ cd Faster-RCNN-with-torchvision/$ sudo pip install -r requirements.txt数据...
Faster-rcnn:我对Faster-RCNN（Pytorch）的实现
05-14
进度提示（已终结） 完成README-工程代码部分：整体代码提交与测试　2018.6.6 完成代码提交　2018.6.6 完成代码测试： 完成README-部分－2018.6.7 完成README-部分工程代码这个程序是基于Pytorch实现Faster-RCNN功能...
faster-rcnn权重文件vgg16-caffe
02-28
faster-rcnn权重文件vgg16-caffe
基于pytorch框架的Faster-RCNN模型python源码+项目说明+数据集.zip
最新发布
04-12
基于pytorch框架的Faster-RCNN模型python源码+项目说明+数据集.zip 基于pytorch框架的Faster-RCNN模型python源码+项目说明+数据集.zip 基于pytorch框架的Faster-RCNN模型python源码+项目说明+数据集.zip 基于pytorch...
faster rcnn网络结构图解
01-07
faster rcnn网络结构图解
py-faster-rcnn-windows-master.zip
11-26
py-faster-rcnn-windows-master.zip ，caffe配置，替换文件夹下的lib
目标检测（八）--Faster R-CNN
工作笔记
03-28 1万+
转自：www.cnblogs.com/dudumiaomiao/p/6560841.html 略删改 R-CNN --> FAST-RCNN --> FASTER-RCNN R-CNN: (1)输入测试图像； (2)利用selective search 算法在图像中从上到下提取2000个左右的Region Proposal； (3)将每个Region Proposal缩放(w...
FasterRCNN模型结构——pytorch实现
Peach_____的博客
01-19 633
【代码】FasterRCNN模型结构——pytorch实现。
【代码阅读】Faster RCNN 网络结构详解
麒麒哈尔的博客
09-14 1万+
这里写自定义目录标题Faster RCNNImage Pro-Processing主干网络 Faster RCNN作为two stage目标检测方法的代表作，其中有很多模块非常经典，在后续的two stage的模型中有广泛的应用。而且在3D Object Detection的问题中，也是如此。这就回顾一下Faster RCNN的网络结构。 这里推荐一个博客，http://www.telesens....
Faster RCNN基本结构和各部分原理详解
就随便写写....
11-10 3247
Faster RCNN的基本结构
ATOM 网络模型（ResNet50）
surprising
10-16 778
ATOMnet( (feature_extractor): ResNet( (conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False) (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_s...
CNN结构：Windows使用FasterRCNN-C++版本
alppkk4545的博客
11-28 341
参考文章：Windows下VS2013 C++编译测试faster-rcnn。 本文与作者的所写方法有些许不同，欲速则不达，没有按照作者的推荐方法，绕了个弯弯。 Windows版本纯C++版本的FasterRCNN比较难找,且懒得翻译Matlab版本代码，暂时可用的是这个项目：Windows下V...
Faster-RCNN的结构
05-23
Faster-RCNN是一种经典的目标检测算法，其结构主要包括以下几个部分： 1. 特征提取网络：使用卷积神经网络（CNN）来提取输入图像的特征，通常使用预训练的网络如VGG、ResNet等。 2. 区域提取网络：利用Region Proposal Network（RPN）来生成候选区域，即可能包含目标的区域。 3. ROI Pooling层：将RPN生成的候选区域映射到特征图上，并将其裁剪成固定大小的特征图块。 4. 目标分类网络：对于每个候选区域，将其输入到全连接层中进行目标分类。 5. 目标框回归网络：对于每个候选区域，将其输入到全连接层中进行目标框的回归，即调整候选框的位置和大小。 Faster-RCNN的主要思想是利用RPN来生成候选区域，而不是使用传统的滑动窗口方法。这样可以大大减少计算量，并且使得检测速度更快。同时，Faster-RCNN采用了端到端的训练方法，可以直接从原始图像中学习目标检测模型，而无需手工设计特征。

“相关推荐”对你有帮助么？

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值