adding identity block (dabdfa1a) · Commits · 钟慕尧 / deepchem

deepchem/models/tensorgraph/models/resnet50.py

+66 −3

Original line number	Diff line number	Diff line
		@@ -10,7 +10,7 @@ import tensorflow as tf
		import deepchem as dc
		from deepchem.models import Sequential
		from deepchem.models.tensorgraph.layers import Conv2D, MaxPool2D, Conv2DTranspose, Concat, Feature, Label, BatchNorm
		from deepchem.models.tensorgraph.layers import SoftMaxCrossEntropy, ReduceMean, SoftMax, ReLU, Add
		from deepchem.models.tensorgraph.layers import SoftMaxCrossEntropy, ReduceMean, SoftMax, ReLU, Add, Flatten, Dense
		from deepchem.models import TensorGraph


		@@ -29,6 +29,43 @@ class ResNet50(TensorGraph):
		specifies number of classes
		"""

		def identity_block(input, kernel_size, filters):
		filters1, filters2, filters3 = filters

		output = Conv2D(
		num_outputs=filters1,
		kernel_size=1,
		stride=strides,
		activation='linear',
		padding='same',
		in_layers=[input])
		output = BatchNorm(in_layers=[output])
		output = ReLU(output)

		output = Conv2D(
		num_outputs=filters2,
		kernel_size=kernel_size,
		stride=strides,
		activation='linear',
		padding='same',
		in_layers=[input])
		output = BatchNorm(in_layers=[output])
		output = ReLU(output)

		output = Conv2D(
		num_outputs=filters3,
		kernel_size=1,
		stride=strides,
		activation='linear',
		padding='same',
		in_layers=[input])
		output = BatchNorm(in_layers=[output])

		output = Add(in_layers=[output, input])
		output = ReLU(output)

		return output

		def conv_block(input, kernel_size, filters, strides=2):
		filters1, filters2, filters3 = filters
		output = Conv2D(
		@@ -54,7 +91,6 @@ class ResNet50(TensorGraph):
		output = Conv2D(
		num_outputs=filters3,
		kernel_size=1,
		stride=2,
		activation='linear',
		padding='same',
		in_layers=[output])
		@@ -63,7 +99,6 @@ class ResNet50(TensorGraph):
		shortcut = Conv2D(
		num_outputs=filters3,
		kernel_size=1,
		stride=2,
		activation='linear',
		padding='same',
		in_layers=[input])
		@@ -100,3 +135,31 @@ class ResNet50(TensorGraph):
		pool1 = MaxPool2D(ksize=[1, 2, 2, 1], in_layers=[bn1])

		cb1 = conv_block(pool1, 3, [64, 64, 256], 1)
		id1 = identity_block(cb1, 3, [64, 64, 256])
		id1 = identity_block(id1, 3, [64, 64, 256])

		cb2 = conv_block(id1, 3, [128, 128, 512])
		id2 = identity_block(cb2, 3, [128, 128, 512])
		id2 = identity_block(id2, 3, [128, 128, 512])
		id2 = identity_block(id2, 3, [128, 128, 512])

		cb3 = conv_block(id2, 3, [256, 256, 1024])
		id3 = identity_block(cb3, 3, [256, 256, 1024])
		id3 = identity_block(id3, 3, [256, 256, 1024])
		id3 = identity_block(id3, 3, [256, 256, 1024])
		id3 = identity_block(cb3, 3, [256, 256, 1024])
		id3 = identity_block(id3, 3, [256, 256, 1024])

		cb4 = conv_block(id3, 3, [512, 512, 2048])
		id4 = identity_block(cb4, 3, [512, 512, 2048])
		id4 = identity_block(id4, 3, [512, 512, 2048])

		pool2 = MaxPool2D(ksize=[1, 7, 7, 1], in_layers=[id4])

		flatten = Flatten(in_layers=pool2)
		dense = Dense(classes, in_layers=flatten)

		loss = SoftMaxCrossEntropy(in_layers=[labels, dense])
		loss = ReduceMean(in_layers=[loss])
		self.set_loss(loss)
		self.add_output(dense)

Admin message