/pytorch_tiramisu

FC-DenseNet in PyTorch for Semantic Segmentation

Primary LanguageJupyter NotebookMIT LicenseMIT

One Hundred Layers Tiramisu

PyTorch implementation of The One Hundred Layers Tiramisu: Fully Convolutional DenseNets for Semantic Segmentation.

Tiramisu combines DensetNet and U-Net for high performance semantic segmentation. In this repository, we attempt to replicate the authors' results on the CamVid dataset.

Setup

Requires Anaconda for Python3 installed.

conda create --name tiramisu python=3.6
source activate tiramisu
conda install pytorch torchvision -c pytorch

The train.ipynb notebook shows a basic train/test workflow.

Dataset

Download

Specs

  • Training: 367 frames
  • Validation: 101 frames
  • TestSet: 233 frames
  • Dimensions: 360x480
  • Classes: 11 (+1 background)

Architecture

Tiramisu adopts the UNet design with downsampling, bottleneck, and upsampling paths and skip connections. It replaces convolution and max pooling layers with Dense blocks from the DenseNet architecture. Dense blocks contain residual connections like in ResNet except they concatenate, rather than sum, prior feature maps.

Layers

FCDenseNet103

Authors' Results

Authors Results on CamVid

Authors Results on CamVid

Our Results

FCDenseNet67

We trained for 670 epochs (224x224 crops) with 100 epochs fine-tuning (full-size images). The authors mention "global accuracy" of 90.8 for FC-DenseNet67 on Camvid, compared to our 86.8. If we exclude the 'background' class, accuracy increases to ~89%. We think the authors did this, but haven't confirmed.

Dataset     Loss Accuracy
Validation .209 92.5
Testset .435 86.8

Our Results on CamVid

FCDenseNet103

We trained for 874 epochs with 50 epochs fine-tuning.

Dataset     Loss Accuracy
Validation .178 92.8
Testset .441 86.6

Our Results on CamVid

Predictions

Our Results on CamVid

Training

Hyperparameters

  • WeightInitialization = HeUniform
  • Optimizer = RMSProp
  • LR = .001 with exponential decay of 0.995 after each epoch
  • Data Augmentation = Random Crops, Vertical Flips
  • ValidationSet with early stopping based on IoU or MeanAccuracy with patience of 100 (50 during finetuning)
  • WeightDecay = .0001
  • Finetune with full-size images, LR = .0001
  • Dropout = 0.2
  • BatchNorm "we use current batch stats at training, validation, and test time"

References and Links