yejg2017 / Deeplab-pytorch-1

Deeplab for semantic segmentation tasks

[PYTORCH] Deeplab

Introduction

Here is my pytorch implementation of the model described in the paper DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs paper.

How to use my code

With my code, you can:

• Train your model from scratch
• Train your model with my trained model
• Evaluate test images with either my trained model or yours

Requirements:

• python 3.6
• pytorch 0.4
• opencv (cv2)
• tensorboard
• tensorboardX (This library could be skipped if you do not use SummaryWriter)
• torchvision
• PIL
• numpy

Datasets:

I used 2 different datases: VOC2012 and VOCaugmented (VOC2007 + 2012) Statistics of datasets I used for experiments is shown below

Dataset	#Classes	#Train images	#Validation images
VOC2012	20	5011	1449
VOCaugmented	20	1464	1449

Create a data folder under the repository,

cd {repo_root}
mkdir data

• VOC: Download the voc images and annotations from VOC2007 or VOC2012. Make sure to put the files as the following structure:

  VOCDevkit
  ├── VOC2007
  │   ├── Annotations  
  │   ├── ImageSets
  │   ├── JPEGImages
  │   └── ...
  ├── VOC2012
  │   ├── Annotations  
  │   ├── ImageSets
  │   ├── JPEGImages
  │   └── ...
  └── VOCaugmented
      ├── gt  
      ├── img
      ├── list
      └── ...
  

Note: You need to put ALL images from 2 dataset VOC2007 and VOC2012 into folder VOCdevkit/VOCaugmented/img/

• In my implementation, in every epoch, the model is saved only when its loss is the lowest one so far. You could also use early stopping, which could be triggered by specifying a positive integer value for parameter es_patience, to stop training process when validation loss has not been improved for es_patience epoches.

Trained models

You could find all trained models I have trained in Deeplab trained models

Training

I provide my pre-trained model name vietnh_trained_deeplab_voc. You could put it in the folder trained_models/, and load it before training your new model, for faster convergence.

If you want to train a new model, you could run:

• python3 train_voc.py --dataset dataset: For example, python train_voc.py --dataset augmentedvoc

Test

By default, my test script will load trained model from folder trained_models/. You of course could change it to other folder which contains your trained model(s).

I provide 2 different test scripts:

If you want to test a trained model with a standard VOC dataset, you could run:

• python3 test_voc.py --year year: For example, python3 test_voc.py --year 2012

If you want to test a model with some images, you could put them into the folder test_images/, then run:

• python3 test_voc_single_images.py --input --output path/to/output/folder: For example, python3 test_voc_single_images.py --output predictions. For easy comparison, not only output images are created, but input images are also copied to output folder

Experiments:

I trained models in 2 machines, one with NVIDIA TITAN X 12gb GPU and the other with NVIDIA quadro 6000 24gb GPU.

The training/test loss curves for each experiment are shown below:

• VOC2012
• VOCaugmented

Results

Some output predictions for experiments for each dataset are shown below:

• VOC2012

• VOCaugmented