50 lines
1.6 KiB
Python
50 lines
1.6 KiB
Python
# Copyright (c) 2019, NVIDIA CORPORATION. All rights reserved.
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
|
|
""" Model construction utils
|
|
|
|
This module provides a convenient way to create different topologies
|
|
based around UNet.
|
|
|
|
"""
|
|
import tensorflow as tf
|
|
from model.layers import output_block, upsample_block, bottleneck, downsample_block, input_block
|
|
|
|
|
|
def unet_v1(features, mode):
|
|
""" U-Net: Convolutional Networks for Biomedical Image Segmentation
|
|
|
|
Source:
|
|
https://arxiv.org/pdf/1505.04597
|
|
|
|
"""
|
|
|
|
skip_connections = []
|
|
|
|
out, skip = input_block(features, filters=64)
|
|
|
|
skip_connections.append(skip)
|
|
|
|
for idx, filters in enumerate([128, 256, 512]):
|
|
out, skip = downsample_block(out, filters=filters, idx=idx)
|
|
skip_connections.append(skip)
|
|
|
|
out = bottleneck(out, filters=1024, mode=mode)
|
|
|
|
for idx, filters in enumerate([512, 256, 128]):
|
|
out = upsample_block(out,
|
|
residual_input=skip_connections.pop(),
|
|
filters=filters,
|
|
idx=idx)
|
|
return output_block(out, residual_input=skip_connections.pop(), filters=64, n_classes=2)
|