This repository contains the implementation of "The One Hundred Layers Tiramisu: Fully Convolutional DenseNets for Semantic Segmentation" in pytorch.
- The decoder part of this implementation is bit different from that of the paper.
from tiramisu import Tiramisu_Segmentation
net = Tiramisu_Segmentation(layer_tiramisu=103,nclasses=1,input_features=1,growth_rate=16)
"""
Arguments :
layer_tiramisu - 57, 47 or 103
input_features - input image channels
nclasses - number of classes
growth_rate - growth rate (filters to begin with for convolution - generally 16)
"""
-
Example 1
-
Example 2