Code for Detection Of Road Damage Using Faster Regional-Convolutional Neural Network Method paper
This is tensorflow Faster-RCNN implementation from scratch supporting to the batch processing. All methods are tried to be created in the simplest way for easy understanding. Most of the operations performed during the implementation were carried out as described in the paper and tf-rpn repository.
It's implemented and tested with tensorflow 2.8, and Python 3.7.9.
VGG-16, ResNet-50, and MobilenetV2 are supported in this project
This project use 300×300 size for input. Please change the feature_map_shape according to your input shape size.
Check train_utils.py to change feature_map_shape or input shape size (img_size)
To check value for feature_map_shape you can run python in terminal and get the summary, to make it simpler I'll show you how to do it.
from tensorflow.keras.applications.vgg16 import VGG16
a = VGG16(include_top = False, input_shape = [300, 300, 3])
a.summary()And so the output will be like this
View Output
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_1 (InputLayer) [(None, 300, 300, 3)] 0
block1_conv1 (Conv2D) (None, 300, 300, 64) 1792
block1_conv2 (Conv2D) (None, 300, 300, 64) 36928
block1_pool (MaxPooling2D) (None, 150, 150, 64) 0
block2_conv1 (Conv2D) (None, 150, 150, 128) 73856
block2_conv2 (Conv2D) (None, 150, 150, 128) 147584
block2_pool (MaxPooling2D) (None, 75, 75, 128) 0
block3_conv1 (Conv2D) (None, 75, 75, 256) 295168
block3_conv2 (Conv2D) (None, 75, 75, 256) 590080
block3_conv3 (Conv2D) (None, 75, 75, 256) 590080
block3_pool (MaxPooling2D) (None, 37, 37, 256) 0
block4_conv1 (Conv2D) (None, 37, 37, 512) 1180160
block4_conv2 (Conv2D) (None, 37, 37, 512) 2359808
block4_conv3 (Conv2D) (None, 37, 37, 512) 2359808
block4_pool (MaxPooling2D) (None, 18, 18, 512) 0
block5_conv1 (Conv2D) (None, 18, 18, 512) 2359808
block5_conv2 (Conv2D) (None, 18, 18, 512) 2359808
block5_conv3 (Conv2D) (None, 18, 18, 512) 2359808
block5_pool (MaxPooling2D) (None, 9, 9, 512) 0
=================================================================
Total params: 14,714,688
Trainable params: 14,714,688
Non-trainable params: 0
_________________________________________________________________
from tensorflow.keras.applications.resnet50 import ResNet50
a = ResNet50(include_top = False, input_shape = [300, 300, 3])
a.summary()And so the output will be like this
View Output
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input_2 (InputLayer) [(None, 300, 300, 3 0 []
)]
conv1_pad (ZeroPadding2D) (None, 306, 306, 3) 0 ['input_2[0][0]']
conv1_conv (Conv2D) (None, 150, 150, 64 9472 ['conv1_pad[0][0]']
)
conv1_bn (BatchNormalization) (None, 150, 150, 64 256 ['conv1_conv[0][0]']
)
conv1_relu (Activation) (None, 150, 150, 64 0 ['conv1_bn[0][0]']
)
pool1_pad (ZeroPadding2D) (None, 152, 152, 64 0 ['conv1_relu[0][0]']
)
pool1_pool (MaxPooling2D) (None, 75, 75, 64) 0 ['pool1_pad[0][0]']
conv2_block1_1_conv (Conv2D) (None, 75, 75, 64) 4160 ['pool1_pool[0][0]']
conv2_block1_1_bn (BatchNormal (None, 75, 75, 64) 256 ['conv2_block1_1_conv[0][0]']
ization)
conv2_block1_1_relu (Activatio (None, 75, 75, 64) 0 ['conv2_block1_1_bn[0][0]']
n)
conv2_block1_2_conv (Conv2D) (None, 75, 75, 64) 36928 ['conv2_block1_1_relu[0][0]']
conv2_block1_2_bn (BatchNormal (None, 75, 75, 64) 256 ['conv2_block1_2_conv[0][0]']
ization)
conv2_block1_2_relu (Activatio (None, 75, 75, 64) 0 ['conv2_block1_2_bn[0][0]']
n)
conv2_block1_0_conv (Conv2D) (None, 75, 75, 256) 16640 ['pool1_pool[0][0]']
conv2_block1_3_conv (Conv2D) (None, 75, 75, 256) 16640 ['conv2_block1_2_relu[0][0]']
conv2_block1_0_bn (BatchNormal (None, 75, 75, 256) 1024 ['conv2_block1_0_conv[0][0]']
ization)
conv2_block1_3_bn (BatchNormal (None, 75, 75, 256) 1024 ['conv2_block1_3_conv[0][0]']
ization)
conv2_block1_add (Add) (None, 75, 75, 256) 0 ['conv2_block1_0_bn[0][0]',
'conv2_block1_3_bn[0][0]']
conv2_block1_out (Activation) (None, 75, 75, 256) 0 ['conv2_block1_add[0][0]']
conv2_block2_1_conv (Conv2D) (None, 75, 75, 64) 16448 ['conv2_block1_out[0][0]']
conv2_block2_1_bn (BatchNormal (None, 75, 75, 64) 256 ['conv2_block2_1_conv[0][0]']
ization)
conv2_block2_1_relu (Activatio (None, 75, 75, 64) 0 ['conv2_block2_1_bn[0][0]']
n)
conv2_block2_2_conv (Conv2D) (None, 75, 75, 64) 36928 ['conv2_block2_1_relu[0][0]']
conv2_block2_2_bn (BatchNormal (None, 75, 75, 64) 256 ['conv2_block2_2_conv[0][0]']
ization)
conv2_block2_2_relu (Activatio (None, 75, 75, 64) 0 ['conv2_block2_2_bn[0][0]']
n)
conv2_block2_3_conv (Conv2D) (None, 75, 75, 256) 16640 ['conv2_block2_2_relu[0][0]']
conv2_block2_3_bn (BatchNormal (None, 75, 75, 256) 1024 ['conv2_block2_3_conv[0][0]']
ization)
conv2_block2_add (Add) (None, 75, 75, 256) 0 ['conv2_block1_out[0][0]',
'conv2_block2_3_bn[0][0]']
conv2_block2_out (Activation) (None, 75, 75, 256) 0 ['conv2_block2_add[0][0]']
conv2_block3_1_conv (Conv2D) (None, 75, 75, 64) 16448 ['conv2_block2_out[0][0]']
conv2_block3_1_bn (BatchNormal (None, 75, 75, 64) 256 ['conv2_block3_1_conv[0][0]']
ization)
conv2_block3_1_relu (Activatio (None, 75, 75, 64) 0 ['conv2_block3_1_bn[0][0]']
n)
conv2_block3_2_conv (Conv2D) (None, 75, 75, 64) 36928 ['conv2_block3_1_relu[0][0]']
conv2_block3_2_bn (BatchNormal (None, 75, 75, 64) 256 ['conv2_block3_2_conv[0][0]']
ization)
conv2_block3_2_relu (Activatio (None, 75, 75, 64) 0 ['conv2_block3_2_bn[0][0]']
n)
conv2_block3_3_conv (Conv2D) (None, 75, 75, 256) 16640 ['conv2_block3_2_relu[0][0]']
conv2_block3_3_bn (BatchNormal (None, 75, 75, 256) 1024 ['conv2_block3_3_conv[0][0]']
ization)
conv2_block3_add (Add) (None, 75, 75, 256) 0 ['conv2_block2_out[0][0]',
'conv2_block3_3_bn[0][0]']
conv2_block3_out (Activation) (None, 75, 75, 256) 0 ['conv2_block3_add[0][0]']
conv3_block1_1_conv (Conv2D) (None, 38, 38, 128) 32896 ['conv2_block3_out[0][0]']
conv3_block1_1_bn (BatchNormal (None, 38, 38, 128) 512 ['conv3_block1_1_conv[0][0]']
ization)
conv3_block1_1_relu (Activatio (None, 38, 38, 128) 0 ['conv3_block1_1_bn[0][0]']
n)
conv3_block1_2_conv (Conv2D) (None, 38, 38, 128) 147584 ['conv3_block1_1_relu[0][0]']
conv3_block1_2_bn (BatchNormal (None, 38, 38, 128) 512 ['conv3_block1_2_conv[0][0]']
ization)
conv3_block1_2_relu (Activatio (None, 38, 38, 128) 0 ['conv3_block1_2_bn[0][0]']
n)
conv3_block1_0_conv (Conv2D) (None, 38, 38, 512) 131584 ['conv2_block3_out[0][0]']
conv3_block1_3_conv (Conv2D) (None, 38, 38, 512) 66048 ['conv3_block1_2_relu[0][0]']
conv3_block1_0_bn (BatchNormal (None, 38, 38, 512) 2048 ['conv3_block1_0_conv[0][0]']
ization)
conv3_block1_3_bn (BatchNormal (None, 38, 38, 512) 2048 ['conv3_block1_3_conv[0][0]']
ization)
conv3_block1_add (Add) (None, 38, 38, 512) 0 ['conv3_block1_0_bn[0][0]',
'conv3_block1_3_bn[0][0]']
conv3_block1_out (Activation) (None, 38, 38, 512) 0 ['conv3_block1_add[0][0]']
conv3_block2_1_conv (Conv2D) (None, 38, 38, 128) 65664 ['conv3_block1_out[0][0]']
conv3_block2_1_bn (BatchNormal (None, 38, 38, 128) 512 ['conv3_block2_1_conv[0][0]']
ization)
conv3_block2_1_relu (Activatio (None, 38, 38, 128) 0 ['conv3_block2_1_bn[0][0]']
n)
conv3_block2_2_conv (Conv2D) (None, 38, 38, 128) 147584 ['conv3_block2_1_relu[0][0]']
conv3_block2_2_bn (BatchNormal (None, 38, 38, 128) 512 ['conv3_block2_2_conv[0][0]']
ization)
conv3_block2_2_relu (Activatio (None, 38, 38, 128) 0 ['conv3_block2_2_bn[0][0]']
n)
conv3_block2_3_conv (Conv2D) (None, 38, 38, 512) 66048 ['conv3_block2_2_relu[0][0]']
conv3_block2_3_bn (BatchNormal (None, 38, 38, 512) 2048 ['conv3_block2_3_conv[0][0]']
ization)
conv3_block2_add (Add) (None, 38, 38, 512) 0 ['conv3_block1_out[0][0]',
'conv3_block2_3_bn[0][0]']
conv3_block2_out (Activation) (None, 38, 38, 512) 0 ['conv3_block2_add[0][0]']
conv3_block3_1_conv (Conv2D) (None, 38, 38, 128) 65664 ['conv3_block2_out[0][0]']
conv3_block3_1_bn (BatchNormal (None, 38, 38, 128) 512 ['conv3_block3_1_conv[0][0]']
ization)
conv3_block3_1_relu (Activatio (None, 38, 38, 128) 0 ['conv3_block3_1_bn[0][0]']
n)
conv3_block3_2_conv (Conv2D) (None, 38, 38, 128) 147584 ['conv3_block3_1_relu[0][0]']
conv3_block3_2_bn (BatchNormal (None, 38, 38, 128) 512 ['conv3_block3_2_conv[0][0]']
ization)
conv3_block3_2_relu (Activatio (None, 38, 38, 128) 0 ['conv3_block3_2_bn[0][0]']
n)
conv3_block3_3_conv (Conv2D) (None, 38, 38, 512) 66048 ['conv3_block3_2_relu[0][0]']
conv3_block3_3_bn (BatchNormal (None, 38, 38, 512) 2048 ['conv3_block3_3_conv[0][0]']
ization)
conv3_block3_add (Add) (None, 38, 38, 512) 0 ['conv3_block2_out[0][0]',
'conv3_block3_3_bn[0][0]']
conv3_block3_out (Activation) (None, 38, 38, 512) 0 ['conv3_block3_add[0][0]']
conv3_block4_1_conv (Conv2D) (None, 38, 38, 128) 65664 ['conv3_block3_out[0][0]']
conv3_block4_1_bn (BatchNormal (None, 38, 38, 128) 512 ['conv3_block4_1_conv[0][0]']
ization)
conv3_block4_1_relu (Activatio (None, 38, 38, 128) 0 ['conv3_block4_1_bn[0][0]']
n)
conv3_block4_2_conv (Conv2D) (None, 38, 38, 128) 147584 ['conv3_block4_1_relu[0][0]']
conv3_block4_2_bn (BatchNormal (None, 38, 38, 128) 512 ['conv3_block4_2_conv[0][0]']
ization)
conv3_block4_2_relu (Activatio (None, 38, 38, 128) 0 ['conv3_block4_2_bn[0][0]']
n)
conv3_block4_3_conv (Conv2D) (None, 38, 38, 512) 66048 ['conv3_block4_2_relu[0][0]']
conv3_block4_3_bn (BatchNormal (None, 38, 38, 512) 2048 ['conv3_block4_3_conv[0][0]']
ization)
conv3_block4_add (Add) (None, 38, 38, 512) 0 ['conv3_block3_out[0][0]',
'conv3_block4_3_bn[0][0]']
conv3_block4_out (Activation) (None, 38, 38, 512) 0 ['conv3_block4_add[0][0]']
conv4_block1_1_conv (Conv2D) (None, 19, 19, 256) 131328 ['conv3_block4_out[0][0]']
conv4_block1_1_bn (BatchNormal (None, 19, 19, 256) 1024 ['conv4_block1_1_conv[0][0]']
ization)
conv4_block1_1_relu (Activatio (None, 19, 19, 256) 0 ['conv4_block1_1_bn[0][0]']
n)
conv4_block1_2_conv (Conv2D) (None, 19, 19, 256) 590080 ['conv4_block1_1_relu[0][0]']
conv4_block1_2_bn (BatchNormal (None, 19, 19, 256) 1024 ['conv4_block1_2_conv[0][0]']
ization)
conv4_block1_2_relu (Activatio (None, 19, 19, 256) 0 ['conv4_block1_2_bn[0][0]']
n)
conv4_block1_0_conv (Conv2D) (None, 19, 19, 1024 525312 ['conv3_block4_out[0][0]']
)
conv4_block1_3_conv (Conv2D) (None, 19, 19, 1024 263168 ['conv4_block1_2_relu[0][0]']
)
conv4_block1_0_bn (BatchNormal (None, 19, 19, 1024 4096 ['conv4_block1_0_conv[0][0]']
ization) )
conv4_block1_3_bn (BatchNormal (None, 19, 19, 1024 4096 ['conv4_block1_3_conv[0][0]']
ization) )
conv4_block1_add (Add) (None, 19, 19, 1024 0 ['conv4_block1_0_bn[0][0]',
) 'conv4_block1_3_bn[0][0]']
conv4_block1_out (Activation) (None, 19, 19, 1024 0 ['conv4_block1_add[0][0]']
)
conv4_block2_1_conv (Conv2D) (None, 19, 19, 256) 262400 ['conv4_block1_out[0][0]']
conv4_block2_1_bn (BatchNormal (None, 19, 19, 256) 1024 ['conv4_block2_1_conv[0][0]']
ization)
conv4_block2_1_relu (Activatio (None, 19, 19, 256) 0 ['conv4_block2_1_bn[0][0]']
n)
conv4_block2_2_conv (Conv2D) (None, 19, 19, 256) 590080 ['conv4_block2_1_relu[0][0]']
conv4_block2_2_bn (BatchNormal (None, 19, 19, 256) 1024 ['conv4_block2_2_conv[0][0]']
ization)
conv4_block2_2_relu (Activatio (None, 19, 19, 256) 0 ['conv4_block2_2_bn[0][0]']
n)
conv4_block2_3_conv (Conv2D) (None, 19, 19, 1024 263168 ['conv4_block2_2_relu[0][0]']
)
conv4_block2_3_bn (BatchNormal (None, 19, 19, 1024 4096 ['conv4_block2_3_conv[0][0]']
ization) )
conv4_block2_add (Add) (None, 19, 19, 1024 0 ['conv4_block1_out[0][0]',
) 'conv4_block2_3_bn[0][0]']
conv4_block2_out (Activation) (None, 19, 19, 1024 0 ['conv4_block2_add[0][0]']
)
conv4_block3_1_conv (Conv2D) (None, 19, 19, 256) 262400 ['conv4_block2_out[0][0]']
conv4_block3_1_bn (BatchNormal (None, 19, 19, 256) 1024 ['conv4_block3_1_conv[0][0]']
ization)
conv4_block3_1_relu (Activatio (None, 19, 19, 256) 0 ['conv4_block3_1_bn[0][0]']
n)
conv4_block3_2_conv (Conv2D) (None, 19, 19, 256) 590080 ['conv4_block3_1_relu[0][0]']
conv4_block3_2_bn (BatchNormal (None, 19, 19, 256) 1024 ['conv4_block3_2_conv[0][0]']
ization)
conv4_block3_2_relu (Activatio (None, 19, 19, 256) 0 ['conv4_block3_2_bn[0][0]']
n)
conv4_block3_3_conv (Conv2D) (None, 19, 19, 1024 263168 ['conv4_block3_2_relu[0][0]']
)
conv4_block3_3_bn (BatchNormal (None, 19, 19, 1024 4096 ['conv4_block3_3_conv[0][0]']
ization) )
conv4_block3_add (Add) (None, 19, 19, 1024 0 ['conv4_block2_out[0][0]',
) 'conv4_block3_3_bn[0][0]']
conv4_block3_out (Activation) (None, 19, 19, 1024 0 ['conv4_block3_add[0][0]']
)
conv4_block4_1_conv (Conv2D) (None, 19, 19, 256) 262400 ['conv4_block3_out[0][0]']
conv4_block4_1_bn (BatchNormal (None, 19, 19, 256) 1024 ['conv4_block4_1_conv[0][0]']
ization)
conv4_block4_1_relu (Activatio (None, 19, 19, 256) 0 ['conv4_block4_1_bn[0][0]']
n)
conv4_block4_2_conv (Conv2D) (None, 19, 19, 256) 590080 ['conv4_block4_1_relu[0][0]']
conv4_block4_2_bn (BatchNormal (None, 19, 19, 256) 1024 ['conv4_block4_2_conv[0][0]']
ization)
conv4_block4_2_relu (Activatio (None, 19, 19, 256) 0 ['conv4_block4_2_bn[0][0]']
n)
conv4_block4_3_conv (Conv2D) (None, 19, 19, 1024 263168 ['conv4_block4_2_relu[0][0]']
)
conv4_block4_3_bn (BatchNormal (None, 19, 19, 1024 4096 ['conv4_block4_3_conv[0][0]']
ization) )
conv4_block4_add (Add) (None, 19, 19, 1024 0 ['conv4_block3_out[0][0]',
) 'conv4_block4_3_bn[0][0]']
conv4_block4_out (Activation) (None, 19, 19, 1024 0 ['conv4_block4_add[0][0]']
)
conv4_block5_1_conv (Conv2D) (None, 19, 19, 256) 262400 ['conv4_block4_out[0][0]']
conv4_block5_1_bn (BatchNormal (None, 19, 19, 256) 1024 ['conv4_block5_1_conv[0][0]']
ization)
conv4_block5_1_relu (Activatio (None, 19, 19, 256) 0 ['conv4_block5_1_bn[0][0]']
n)
conv4_block5_2_conv (Conv2D) (None, 19, 19, 256) 590080 ['conv4_block5_1_relu[0][0]']
conv4_block5_2_bn (BatchNormal (None, 19, 19, 256) 1024 ['conv4_block5_2_conv[0][0]']
ization)
conv4_block5_2_relu (Activatio (None, 19, 19, 256) 0 ['conv4_block5_2_bn[0][0]']
n)
conv4_block5_3_conv (Conv2D) (None, 19, 19, 1024 263168 ['conv4_block5_2_relu[0][0]']
)
conv4_block5_3_bn (BatchNormal (None, 19, 19, 1024 4096 ['conv4_block5_3_conv[0][0]']
ization) )
conv4_block5_add (Add) (None, 19, 19, 1024 0 ['conv4_block4_out[0][0]',
) 'conv4_block5_3_bn[0][0]']
conv4_block5_out (Activation) (None, 19, 19, 1024 0 ['conv4_block5_add[0][0]']
)
conv4_block6_1_conv (Conv2D) (None, 19, 19, 256) 262400 ['conv4_block5_out[0][0]']
conv4_block6_1_bn (BatchNormal (None, 19, 19, 256) 1024 ['conv4_block6_1_conv[0][0]']
ization)
conv4_block6_1_relu (Activatio (None, 19, 19, 256) 0 ['conv4_block6_1_bn[0][0]']
n)
conv4_block6_2_conv (Conv2D) (None, 19, 19, 256) 590080 ['conv4_block6_1_relu[0][0]']
conv4_block6_2_bn (BatchNormal (None, 19, 19, 256) 1024 ['conv4_block6_2_conv[0][0]']
ization)
conv4_block6_2_relu (Activatio (None, 19, 19, 256) 0 ['conv4_block6_2_bn[0][0]']
n)
conv4_block6_3_conv (Conv2D) (None, 19, 19, 1024 263168 ['conv4_block6_2_relu[0][0]']
)
conv4_block6_3_bn (BatchNormal (None, 19, 19, 1024 4096 ['conv4_block6_3_conv[0][0]']
ization) )
conv4_block6_add (Add) (None, 19, 19, 1024 0 ['conv4_block5_out[0][0]',
) 'conv4_block6_3_bn[0][0]']
conv4_block6_out (Activation) (None, 19, 19, 1024 0 ['conv4_block6_add[0][0]']
)
conv5_block1_1_conv (Conv2D) (None, 10, 10, 512) 524800 ['conv4_block6_out[0][0]']
conv5_block1_1_bn (BatchNormal (None, 10, 10, 512) 2048 ['conv5_block1_1_conv[0][0]']
ization)
conv5_block1_1_relu (Activatio (None, 10, 10, 512) 0 ['conv5_block1_1_bn[0][0]']
n)
conv5_block1_2_conv (Conv2D) (None, 10, 10, 512) 2359808 ['conv5_block1_1_relu[0][0]']
conv5_block1_2_bn (BatchNormal (None, 10, 10, 512) 2048 ['conv5_block1_2_conv[0][0]']
ization)
conv5_block1_2_relu (Activatio (None, 10, 10, 512) 0 ['conv5_block1_2_bn[0][0]']
n)
conv5_block1_0_conv (Conv2D) (None, 10, 10, 2048 2099200 ['conv4_block6_out[0][0]']
)
conv5_block1_3_conv (Conv2D) (None, 10, 10, 2048 1050624 ['conv5_block1_2_relu[0][0]']
)
conv5_block1_0_bn (BatchNormal (None, 10, 10, 2048 8192 ['conv5_block1_0_conv[0][0]']
ization) )
conv5_block1_3_bn (BatchNormal (None, 10, 10, 2048 8192 ['conv5_block1_3_conv[0][0]']
ization) )
conv5_block1_add (Add) (None, 10, 10, 2048 0 ['conv5_block1_0_bn[0][0]',
) 'conv5_block1_3_bn[0][0]']
conv5_block1_out (Activation) (None, 10, 10, 2048 0 ['conv5_block1_add[0][0]']
)
conv5_block2_1_conv (Conv2D) (None, 10, 10, 512) 1049088 ['conv5_block1_out[0][0]']
conv5_block2_1_bn (BatchNormal (None, 10, 10, 512) 2048 ['conv5_block2_1_conv[0][0]']
ization)
conv5_block2_1_relu (Activatio (None, 10, 10, 512) 0 ['conv5_block2_1_bn[0][0]']
n)
conv5_block2_2_conv (Conv2D) (None, 10, 10, 512) 2359808 ['conv5_block2_1_relu[0][0]']
conv5_block2_2_bn (BatchNormal (None, 10, 10, 512) 2048 ['conv5_block2_2_conv[0][0]']
ization)
conv5_block2_2_relu (Activatio (None, 10, 10, 512) 0 ['conv5_block2_2_bn[0][0]']
n)
conv5_block2_3_conv (Conv2D) (None, 10, 10, 2048 1050624 ['conv5_block2_2_relu[0][0]']
)
conv5_block2_3_bn (BatchNormal (None, 10, 10, 2048 8192 ['conv5_block2_3_conv[0][0]']
ization) )
conv5_block2_add (Add) (None, 10, 10, 2048 0 ['conv5_block1_out[0][0]',
) 'conv5_block2_3_bn[0][0]']
conv5_block2_out (Activation) (None, 10, 10, 2048 0 ['conv5_block2_add[0][0]']
)
conv5_block3_1_conv (Conv2D) (None, 10, 10, 512) 1049088 ['conv5_block2_out[0][0]']
conv5_block3_1_bn (BatchNormal (None, 10, 10, 512) 2048 ['conv5_block3_1_conv[0][0]']
ization)
conv5_block3_1_relu (Activatio (None, 10, 10, 512) 0 ['conv5_block3_1_bn[0][0]']
n)
conv5_block3_2_conv (Conv2D) (None, 10, 10, 512) 2359808 ['conv5_block3_1_relu[0][0]']
conv5_block3_2_bn (BatchNormal (None, 10, 10, 512) 2048 ['conv5_block3_2_conv[0][0]']
ization)
conv5_block3_2_relu (Activatio (None, 10, 10, 512) 0 ['conv5_block3_2_bn[0][0]']
n)
conv5_block3_3_conv (Conv2D) (None, 10, 10, 2048 1050624 ['conv5_block3_2_relu[0][0]']
)
conv5_block3_3_bn (BatchNormal (None, 10, 10, 2048 8192 ['conv5_block3_3_conv[0][0]']
ization) )
conv5_block3_add (Add) (None, 10, 10, 2048 0 ['conv5_block2_out[0][0]',
) 'conv5_block3_3_bn[0][0]']
conv5_block3_out (Activation) (None, 10, 10, 2048 0 ['conv5_block3_add[0][0]']
)
==================================================================================================
Total params: 23,587,712
Trainable params: 23,534,592
Non-trainable params: 53,120
__________________________________________________________________________________________________
from tensorflow.keras.applications.mobilenet_v2 import MobileNetV2
a = MobileNetV2(include_top=False, input_shape = [300, 300, 3])
a.summary()Ignore the warning when you initialize MobileNetV2 model.
And so the output will be like this
View Output
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input_1 (InputLayer) [(None, 300, 300, 3 0 []
)]
Conv1 (Conv2D) (None, 150, 150, 32 864 ['input_1[0][0]']
)
bn_Conv1 (BatchNormalization) (None, 150, 150, 32 128 ['Conv1[0][0]']
)
Conv1_relu (ReLU) (None, 150, 150, 32 0 ['bn_Conv1[0][0]']
)
expanded_conv_depthwise (Depth (None, 150, 150, 32 288 ['Conv1_relu[0][0]']
wiseConv2D) )
expanded_conv_depthwise_BN (Ba (None, 150, 150, 32 128 ['expanded_conv_depthwise[0][0]']
tchNormalization) )
expanded_conv_depthwise_relu ( (None, 150, 150, 32 0 ['expanded_conv_depthwise_BN[0][0
ReLU) ) ]']
expanded_conv_project (Conv2D) (None, 150, 150, 16 512 ['expanded_conv_depthwise_relu[0]
) [0]']
expanded_conv_project_BN (Batc (None, 150, 150, 16 64 ['expanded_conv_project[0][0]']
hNormalization) )
block_1_expand (Conv2D) (None, 150, 150, 96 1536 ['expanded_conv_project_BN[0][0]'
) ]
block_1_expand_BN (BatchNormal (None, 150, 150, 96 384 ['block_1_expand[0][0]']
ization) )
block_1_expand_relu (ReLU) (None, 150, 150, 96 0 ['block_1_expand_BN[0][0]']
)
block_1_pad (ZeroPadding2D) (None, 151, 151, 96 0 ['block_1_expand_relu[0][0]']
)
block_1_depthwise (DepthwiseCo (None, 75, 75, 96) 864 ['block_1_pad[0][0]']
nv2D)
block_1_depthwise_BN (BatchNor (None, 75, 75, 96) 384 ['block_1_depthwise[0][0]']
malization)
block_1_depthwise_relu (ReLU) (None, 75, 75, 96) 0 ['block_1_depthwise_BN[0][0]']
block_1_project (Conv2D) (None, 75, 75, 24) 2304 ['block_1_depthwise_relu[0][0]']
block_1_project_BN (BatchNorma (None, 75, 75, 24) 96 ['block_1_project[0][0]']
lization)
block_2_expand (Conv2D) (None, 75, 75, 144) 3456 ['block_1_project_BN[0][0]']
block_2_expand_BN (BatchNormal (None, 75, 75, 144) 576 ['block_2_expand[0][0]']
ization)
block_2_expand_relu (ReLU) (None, 75, 75, 144) 0 ['block_2_expand_BN[0][0]']
block_2_depthwise (DepthwiseCo (None, 75, 75, 144) 1296 ['block_2_expand_relu[0][0]']
nv2D)
block_2_depthwise_BN (BatchNor (None, 75, 75, 144) 576 ['block_2_depthwise[0][0]']
malization)
block_2_depthwise_relu (ReLU) (None, 75, 75, 144) 0 ['block_2_depthwise_BN[0][0]']
block_2_project (Conv2D) (None, 75, 75, 24) 3456 ['block_2_depthwise_relu[0][0]']
block_2_project_BN (BatchNorma (None, 75, 75, 24) 96 ['block_2_project[0][0]']
lization)
block_2_add (Add) (None, 75, 75, 24) 0 ['block_1_project_BN[0][0]',
'block_2_project_BN[0][0]']
block_3_expand (Conv2D) (None, 75, 75, 144) 3456 ['block_2_add[0][0]']
block_3_expand_BN (BatchNormal (None, 75, 75, 144) 576 ['block_3_expand[0][0]']
ization)
block_3_expand_relu (ReLU) (None, 75, 75, 144) 0 ['block_3_expand_BN[0][0]']
block_3_pad (ZeroPadding2D) (None, 77, 77, 144) 0 ['block_3_expand_relu[0][0]']
block_3_depthwise (DepthwiseCo (None, 38, 38, 144) 1296 ['block_3_pad[0][0]']
nv2D)
block_3_depthwise_BN (BatchNor (None, 38, 38, 144) 576 ['block_3_depthwise[0][0]']
malization)
block_3_depthwise_relu (ReLU) (None, 38, 38, 144) 0 ['block_3_depthwise_BN[0][0]']
block_3_project (Conv2D) (None, 38, 38, 32) 4608 ['block_3_depthwise_relu[0][0]']
block_3_project_BN (BatchNorma (None, 38, 38, 32) 128 ['block_3_project[0][0]']
lization)
block_4_expand (Conv2D) (None, 38, 38, 192) 6144 ['block_3_project_BN[0][0]']
block_4_expand_BN (BatchNormal (None, 38, 38, 192) 768 ['block_4_expand[0][0]']
ization)
block_4_expand_relu (ReLU) (None, 38, 38, 192) 0 ['block_4_expand_BN[0][0]']
block_4_depthwise (DepthwiseCo (None, 38, 38, 192) 1728 ['block_4_expand_relu[0][0]']
nv2D)
block_4_depthwise_BN (BatchNor (None, 38, 38, 192) 768 ['block_4_depthwise[0][0]']
malization)
block_4_depthwise_relu (ReLU) (None, 38, 38, 192) 0 ['block_4_depthwise_BN[0][0]']
block_4_project (Conv2D) (None, 38, 38, 32) 6144 ['block_4_depthwise_relu[0][0]']
block_4_project_BN (BatchNorma (None, 38, 38, 32) 128 ['block_4_project[0][0]']
lization)
block_4_add (Add) (None, 38, 38, 32) 0 ['block_3_project_BN[0][0]',
'block_4_project_BN[0][0]']
block_5_expand (Conv2D) (None, 38, 38, 192) 6144 ['block_4_add[0][0]']
block_5_expand_BN (BatchNormal (None, 38, 38, 192) 768 ['block_5_expand[0][0]']
ization)
block_5_expand_relu (ReLU) (None, 38, 38, 192) 0 ['block_5_expand_BN[0][0]']
block_5_depthwise (DepthwiseCo (None, 38, 38, 192) 1728 ['block_5_expand_relu[0][0]']
nv2D)
block_5_depthwise_BN (BatchNor (None, 38, 38, 192) 768 ['block_5_depthwise[0][0]']
malization)
block_5_depthwise_relu (ReLU) (None, 38, 38, 192) 0 ['block_5_depthwise_BN[0][0]']
block_5_project (Conv2D) (None, 38, 38, 32) 6144 ['block_5_depthwise_relu[0][0]']
block_5_project_BN (BatchNorma (None, 38, 38, 32) 128 ['block_5_project[0][0]']
lization)
block_5_add (Add) (None, 38, 38, 32) 0 ['block_4_add[0][0]',
'block_5_project_BN[0][0]']
block_6_expand (Conv2D) (None, 38, 38, 192) 6144 ['block_5_add[0][0]']
block_6_expand_BN (BatchNormal (None, 38, 38, 192) 768 ['block_6_expand[0][0]']
ization)
block_6_expand_relu (ReLU) (None, 38, 38, 192) 0 ['block_6_expand_BN[0][0]']
block_6_pad (ZeroPadding2D) (None, 39, 39, 192) 0 ['block_6_expand_relu[0][0]']
block_6_depthwise (DepthwiseCo (None, 19, 19, 192) 1728 ['block_6_pad[0][0]']
nv2D)
block_6_depthwise_BN (BatchNor (None, 19, 19, 192) 768 ['block_6_depthwise[0][0]']
malization)
block_6_depthwise_relu (ReLU) (None, 19, 19, 192) 0 ['block_6_depthwise_BN[0][0]']
block_6_project (Conv2D) (None, 19, 19, 64) 12288 ['block_6_depthwise_relu[0][0]']
block_6_project_BN (BatchNorma (None, 19, 19, 64) 256 ['block_6_project[0][0]']
lization)
block_7_expand (Conv2D) (None, 19, 19, 384) 24576 ['block_6_project_BN[0][0]']
block_7_expand_BN (BatchNormal (None, 19, 19, 384) 1536 ['block_7_expand[0][0]']
ization)
block_7_expand_relu (ReLU) (None, 19, 19, 384) 0 ['block_7_expand_BN[0][0]']
block_7_depthwise (DepthwiseCo (None, 19, 19, 384) 3456 ['block_7_expand_relu[0][0]']
nv2D)
block_7_depthwise_BN (BatchNor (None, 19, 19, 384) 1536 ['block_7_depthwise[0][0]']
malization)
block_7_depthwise_relu (ReLU) (None, 19, 19, 384) 0 ['block_7_depthwise_BN[0][0]']
block_7_project (Conv2D) (None, 19, 19, 64) 24576 ['block_7_depthwise_relu[0][0]']
block_7_project_BN (BatchNorma (None, 19, 19, 64) 256 ['block_7_project[0][0]']
lization)
block_7_add (Add) (None, 19, 19, 64) 0 ['block_6_project_BN[0][0]',
'block_7_project_BN[0][0]']
block_8_expand (Conv2D) (None, 19, 19, 384) 24576 ['block_7_add[0][0]']
block_8_expand_BN (BatchNormal (None, 19, 19, 384) 1536 ['block_8_expand[0][0]']
ization)
block_8_expand_relu (ReLU) (None, 19, 19, 384) 0 ['block_8_expand_BN[0][0]']
block_8_depthwise (DepthwiseCo (None, 19, 19, 384) 3456 ['block_8_expand_relu[0][0]']
nv2D)
block_8_depthwise_BN (BatchNor (None, 19, 19, 384) 1536 ['block_8_depthwise[0][0]']
malization)
block_8_depthwise_relu (ReLU) (None, 19, 19, 384) 0 ['block_8_depthwise_BN[0][0]']
block_8_project (Conv2D) (None, 19, 19, 64) 24576 ['block_8_depthwise_relu[0][0]']
block_8_project_BN (BatchNorma (None, 19, 19, 64) 256 ['block_8_project[0][0]']
lization)
block_8_add (Add) (None, 19, 19, 64) 0 ['block_7_add[0][0]',
'block_8_project_BN[0][0]']
block_9_expand (Conv2D) (None, 19, 19, 384) 24576 ['block_8_add[0][0]']
block_9_expand_BN (BatchNormal (None, 19, 19, 384) 1536 ['block_9_expand[0][0]']
ization)
block_9_expand_relu (ReLU) (None, 19, 19, 384) 0 ['block_9_expand_BN[0][0]']
block_9_depthwise (DepthwiseCo (None, 19, 19, 384) 3456 ['block_9_expand_relu[0][0]']
nv2D)
block_9_depthwise_BN (BatchNor (None, 19, 19, 384) 1536 ['block_9_depthwise[0][0]']
malization)
block_9_depthwise_relu (ReLU) (None, 19, 19, 384) 0 ['block_9_depthwise_BN[0][0]']
block_9_project (Conv2D) (None, 19, 19, 64) 24576 ['block_9_depthwise_relu[0][0]']
block_9_project_BN (BatchNorma (None, 19, 19, 64) 256 ['block_9_project[0][0]']
lization)
block_9_add (Add) (None, 19, 19, 64) 0 ['block_8_add[0][0]',
'block_9_project_BN[0][0]']
block_10_expand (Conv2D) (None, 19, 19, 384) 24576 ['block_9_add[0][0]']
block_10_expand_BN (BatchNorma (None, 19, 19, 384) 1536 ['block_10_expand[0][0]']
lization)
block_10_expand_relu (ReLU) (None, 19, 19, 384) 0 ['block_10_expand_BN[0][0]']
block_10_depthwise (DepthwiseC (None, 19, 19, 384) 3456 ['block_10_expand_relu[0][0]']
onv2D)
block_10_depthwise_BN (BatchNo (None, 19, 19, 384) 1536 ['block_10_depthwise[0][0]']
rmalization)
block_10_depthwise_relu (ReLU) (None, 19, 19, 384) 0 ['block_10_depthwise_BN[0][0]']
block_10_project (Conv2D) (None, 19, 19, 96) 36864 ['block_10_depthwise_relu[0][0]']
block_10_project_BN (BatchNorm (None, 19, 19, 96) 384 ['block_10_project[0][0]']
alization)
block_11_expand (Conv2D) (None, 19, 19, 576) 55296 ['block_10_project_BN[0][0]']
block_11_expand_BN (BatchNorma (None, 19, 19, 576) 2304 ['block_11_expand[0][0]']
lization)
block_11_expand_relu (ReLU) (None, 19, 19, 576) 0 ['block_11_expand_BN[0][0]']
block_11_depthwise (DepthwiseC (None, 19, 19, 576) 5184 ['block_11_expand_relu[0][0]']
onv2D)
block_11_depthwise_BN (BatchNo (None, 19, 19, 576) 2304 ['block_11_depthwise[0][0]']
rmalization)
block_11_depthwise_relu (ReLU) (None, 19, 19, 576) 0 ['block_11_depthwise_BN[0][0]']
block_11_project (Conv2D) (None, 19, 19, 96) 55296 ['block_11_depthwise_relu[0][0]']
block_11_project_BN (BatchNorm (None, 19, 19, 96) 384 ['block_11_project[0][0]']
alization)
block_11_add (Add) (None, 19, 19, 96) 0 ['block_10_project_BN[0][0]',
'block_11_project_BN[0][0]']
block_12_expand (Conv2D) (None, 19, 19, 576) 55296 ['block_11_add[0][0]']
block_12_expand_BN (BatchNorma (None, 19, 19, 576) 2304 ['block_12_expand[0][0]']
lization)
block_12_expand_relu (ReLU) (None, 19, 19, 576) 0 ['block_12_expand_BN[0][0]']
block_12_depthwise (DepthwiseC (None, 19, 19, 576) 5184 ['block_12_expand_relu[0][0]']
onv2D)
block_12_depthwise_BN (BatchNo (None, 19, 19, 576) 2304 ['block_12_depthwise[0][0]']
rmalization)
block_12_depthwise_relu (ReLU) (None, 19, 19, 576) 0 ['block_12_depthwise_BN[0][0]']
block_12_project (Conv2D) (None, 19, 19, 96) 55296 ['block_12_depthwise_relu[0][0]']
block_12_project_BN (BatchNorm (None, 19, 19, 96) 384 ['block_12_project[0][0]']
alization)
block_12_add (Add) (None, 19, 19, 96) 0 ['block_11_add[0][0]',
'block_12_project_BN[0][0]']
block_13_expand (Conv2D) (None, 19, 19, 576) 55296 ['block_12_add[0][0]']
block_13_expand_BN (BatchNorma (None, 19, 19, 576) 2304 ['block_13_expand[0][0]']
lization)
block_13_expand_relu (ReLU) (None, 19, 19, 576) 0 ['block_13_expand_BN[0][0]']
block_13_pad (ZeroPadding2D) (None, 21, 21, 576) 0 ['block_13_expand_relu[0][0]']
block_13_depthwise (DepthwiseC (None, 10, 10, 576) 5184 ['block_13_pad[0][0]']
onv2D)
block_13_depthwise_BN (BatchNo (None, 10, 10, 576) 2304 ['block_13_depthwise[0][0]']
rmalization)
block_13_depthwise_relu (ReLU) (None, 10, 10, 576) 0 ['block_13_depthwise_BN[0][0]']
block_13_project (Conv2D) (None, 10, 10, 160) 92160 ['block_13_depthwise_relu[0][0]']
block_13_project_BN (BatchNorm (None, 10, 10, 160) 640 ['block_13_project[0][0]']
alization)
block_14_expand (Conv2D) (None, 10, 10, 960) 153600 ['block_13_project_BN[0][0]']
block_14_expand_BN (BatchNorma (None, 10, 10, 960) 3840 ['block_14_expand[0][0]']
lization)
block_14_expand_relu (ReLU) (None, 10, 10, 960) 0 ['block_14_expand_BN[0][0]']
block_14_depthwise (DepthwiseC (None, 10, 10, 960) 8640 ['block_14_expand_relu[0][0]']
onv2D)
block_14_depthwise_BN (BatchNo (None, 10, 10, 960) 3840 ['block_14_depthwise[0][0]']
rmalization)
block_14_depthwise_relu (ReLU) (None, 10, 10, 960) 0 ['block_14_depthwise_BN[0][0]']
block_14_project (Conv2D) (None, 10, 10, 160) 153600 ['block_14_depthwise_relu[0][0]']
block_14_project_BN (BatchNorm (None, 10, 10, 160) 640 ['block_14_project[0][0]']
alization)
block_14_add (Add) (None, 10, 10, 160) 0 ['block_13_project_BN[0][0]',
'block_14_project_BN[0][0]']
block_15_expand (Conv2D) (None, 10, 10, 960) 153600 ['block_14_add[0][0]']
block_15_expand_BN (BatchNorma (None, 10, 10, 960) 3840 ['block_15_expand[0][0]']
lization)
block_15_expand_relu (ReLU) (None, 10, 10, 960) 0 ['block_15_expand_BN[0][0]']
block_15_depthwise (DepthwiseC (None, 10, 10, 960) 8640 ['block_15_expand_relu[0][0]']
onv2D)
block_15_depthwise_BN (BatchNo (None, 10, 10, 960) 3840 ['block_15_depthwise[0][0]']
rmalization)
block_15_depthwise_relu (ReLU) (None, 10, 10, 960) 0 ['block_15_depthwise_BN[0][0]']
block_15_project (Conv2D) (None, 10, 10, 160) 153600 ['block_15_depthwise_relu[0][0]']
block_15_project_BN (BatchNorm (None, 10, 10, 160) 640 ['block_15_project[0][0]']
alization)
block_15_add (Add) (None, 10, 10, 160) 0 ['block_14_add[0][0]',
'block_15_project_BN[0][0]']
block_16_expand (Conv2D) (None, 10, 10, 960) 153600 ['block_15_add[0][0]']
block_16_expand_BN (BatchNorma (None, 10, 10, 960) 3840 ['block_16_expand[0][0]']
lization)
block_16_expand_relu (ReLU) (None, 10, 10, 960) 0 ['block_16_expand_BN[0][0]']
block_16_depthwise (DepthwiseC (None, 10, 10, 960) 8640 ['block_16_expand_relu[0][0]']
onv2D)
block_16_depthwise_BN (BatchNo (None, 10, 10, 960) 3840 ['block_16_depthwise[0][0]']
rmalization)
block_16_depthwise_relu (ReLU) (None, 10, 10, 960) 0 ['block_16_depthwise_BN[0][0]']
block_16_project (Conv2D) (None, 10, 10, 320) 307200 ['block_16_depthwise_relu[0][0]']
block_16_project_BN (BatchNorm (None, 10, 10, 320) 1280 ['block_16_project[0][0]']
alization)
Conv_1 (Conv2D) (None, 10, 10, 1280 409600 ['block_16_project_BN[0][0]']
)
Conv_1_bn (BatchNormalization) (None, 10, 10, 1280 5120 ['Conv_1[0][0]']
)
out_relu (ReLU) (None, 10, 10, 1280 0 ['Conv_1_bn[0][0]']
)
==================================================================================================
Total params: 2,257,984
Trainable params: 2,223,872
Non-trainable params: 34,112
__________________________________________________________________________________________________
This Python project framework forked from FurkanOM TF Faster R-CNN Repository
You can see my dataset here
The Android App called RoDa Detection (Road Damage Detection)
Few notes for this section
-
The camera function is use native camera from device (not from jetpack), some device may occurring an force close error
-
In this project the model wasn't saved in mobile, but it is stored in a server. So to fetch the picture is using Retrofit
-
Put your BASE_URL in your local.properties file, example
BASE_URL="your link here" -
This Mobile use MVVM architecture