This is an implementation of the Fully-Convolutional DenseNet for Semantic Segmentation from the paper:
The One Hundred Layers Tiramisu: Fully Convolutional DenseNets for Semantic Segmentation
-
FC_DenseNet_Tiramisu.py: The model file. Use this to build your custom FC-DenseNet models, or one from the original paper
-
main.py: Training and Testing on the CamVid dataset
-
helper.py: Quick helper functions for data preparation and visualization
-
utils.py: Utilities for printing, debugging, and testing
-
CamVid: The CamVid datatset for Semantic Segmentation
-
checkpoints: Checkpoint files for each epoch during training
-
Test: Test results including images and per-class accuracies
This project has the following dependencies:
-
Numpy
sudo pip install numpy -
OpenCV Python
sudo apt-get install python-opencv -
TensorFlow
sudo pip install --upgrade tensorflow-gpu
In training, I used a batch size of 1 and image size of 352x480. The following results are for the FC-DenseNet103 model trained for 250 epochs. I used RMSProp with learning rate 0.001 and decay 0.995. Note that the checkpoint files are not uploaded to this repository since they are too big for GitHub (greater than 100 MB)
| Class | Original Accuracy | My Accuracy |
|---|---|---|
| Sky | 93.0 | 94.1 |
| Building | 83.0 | 81.2 |
| Pole | 37.8 | 38.3 |
| Road | 94.5 | 97.5 |
| Pavement | 82.2 | 87.9 |
| Tree | 77.3 | 75.5 |
| SignSymbol | 43.9 | 49.7 |
| Fence | 37.1 | 69.0 |
| Car | 77.3 | 87.0 |
| Pedestrian | 59.6 | 60.3 |
| Bicyclist | 50.5 | 75.3 |
| Unlabelled | N/A | 40.9 |
| Global | 91.5 | 86.0 |
| Loss vs Epochs | Val. Acc. vs Epochs |
|---|---|
 |
 |
| Original | GT | Result |
|---|---|---|
 |
 |
 |