DenseNet implemented in TensorFlow
This repository implements the paper: Densely Connected Convolutional Networks and is specifically designed for easy integration.
Other implementations can be found here
import densenet
with slim.arg_scope(densenet.densenet_arg_scope()):
# DenseNet with bottleneck convolution and feature compression
net, end_points = densenet.densenet_bc_k12_l40(inputs, 10, is_training=False)
# OR without it
net, end_points = densenet.densenet_k12_l40(inputs, 10, is_training=False)