Semantic Instance Segmentation with a Discriminative Loss Function

This repository implements Semantic Instance Segmentation with a Discriminative Loss Function. However, it contains some enhancements.

• Reference paper does not predict semantic segmentation mask, instead it uses ground-truth semantic segmentation mask. This code predicts semantic segmentation mask, similar to Towards End-to-End Lane Detection: an Instance Segmentation Approach.
• Reference paper predicts the number of instances implicity. It predicts embeddings for instances and predicts the number of instances as a result of clustering. Instead, this code predicts the number of instances as an output of network.
• This code uses Spectral Clustering; however, reference paper uses "a fast variant of the mean-shift algorithm".
• Reference paper uses a segmentation network based on ResNet-38. Instead, this code uses ReSeg with skip-connections based on first seven convolutional layers of VGG16 as segmentation network.

---

In prediction phase, network inputs an image and outputs a semantic segmentation mask, the number of instances and embeddings for all pixels in the image. Then, foreground embeddings (which correspond to instances) are selected using semantic segmentation mask and foreground embeddings are clustered into "the number of instances" groups using spectral clustering.

Installation

• Clone this repository : git clone --recursive https://github.com/Wizaron/instance-segmentation-pytorch.git
• Download and install Anaconda or Miniconda
• Create a conda environment : conda env create -f instance-segmentation-pytorch/code/conda_environment.yml

Data

CVPPP

• Download CVPPP dataset and extract downloaded zip file (CVPPP2017_LSC_training.zip) to instance-segmentation-pytorch/data/raw/CVPPP/
• This work uses A1 subset of the dataset.

Cityscapes

• Download Cityscapes dataset and extract downloaded zip files (gtFine_trainvaltest.zip and leftImg8bit_trainvaltest.zip) to instance-segmentation-pytorch/data/raw/cityscapes/

Code Structure

• code: Codes for training and evaluation.
  • lib
    • lib/cvppp_arch.py: Defines network architecture for CVPPP dataset.
    • lib/cityscapes_arch.py: Defines network architecture for cityscapes dataset.
    • lib/model.py: Defines model (optimization, criterion, fit, predict, test, etc.).
    • lib/dataset.py: Data loading, augmentation, minibatching procedures.
    • lib/preprocess.py, lib/utils: Augmentation methods.
    • lib/prediction.py: Prediction module.
    • lib/losses/dice.py: Dice loss for foreground semantic segmentation.
    • lib/losses/discriminative.py: Discriminative loss for instance segmentation.
  • settings
    • settings/CVPPP/data_settings.py, settings/cityscapes/data_settings.py: Defines settings about data.
    • settings/CVPPP/model_settings.py, settings/cityscapes/model_settings.py: Defines settings about model (hyper-parameters).
    • settings/CVPPP/training_settings.py, settings/cityscapes/training_settings.py: Defines settings for training (optimization method, weight decay, augmentation, etc.).
  • train.py: Training script.
  • pred.py: Prediction script for single image.
  • pred_list.py: Prediction scripts for a list of images.
  • evaluate.py: Evaluation script. Calculates SBD (symmetric best dice), |DiC| (absolute difference in count) and Foreground Dice (Dice score for semantic segmentation) as defined in the paper.
• data: Stores data and scripts to prepare dataset for training and evaluation.
  • metadata/CVPPP, metadata/cityscapes: Stores metadata; such as, training, validation and test splits, image shapes etc.
  • processed/CVPPP, processed/cityscapes: Stores processed form of the data.
  • raw/CVPPP, raw/cityscapes: Stores raw form of the data.
  • scripts: Stores scripts to prepare dataset.
    • scripts/CVPPP: For CVPPP dataset.
      • scripts/CVPPP/1-create_annotations.py: Saves annotations as a numpy array to processed/CVPPP/semantic-annotations/ and processed/CVPPP/instance-annotations.
      • scripts/CVPPP/1-remove_alpha.sh: Removes alpha channels from images. (In order to run this script, imagemagick should be installed.).
      • scripts/CVPPP/2-get_image_means-stds.py: Calculates and prints channel-wise means and standard deviations from training subset.
      • scripts/CVPPP/2-get_image_shapes.py: Saves image shapes to metadata/CVPPP/image_shapes.txt.
      • scripts/CVPPP/2-get_number_of_instances.py: Saves the number of instances in each image to metadata/CVPPP/number_of_instances.txt.
      • scripts/CVPPP/2-get_image_paths.py: Saves image paths to metadata/CVPPP/training_image_paths.txt, metadata/CVPPP/validation_image_paths.txt
      • scripts/CVPPP/3-create_dataset.py: Creates an lmdb dataset to processed/CVPPP/lmdb/.
    • scripts/cityscapes: For cityscapes dataset.
      • scripts/cityscapes/1-create-annotations.py: Saves annotations as a numpy array to processed/cityscapes/semantic-annotations and processed/cityscapes/instance-annotations, saves the number of instances in each image to metadata/cityscapes/number_of_instances.txt and save subset lists to metadata/cityscapes/training.lst and metadata/cityscapes/validation.lst.
      • scripts/cityscapes/2-get_image_paths.py: Saves image paths to metadata/cityscapes/training_image_paths.txt, metadata/cityscapes/validation_image_paths.txt
      • scrits/cityscapes/3-create_dataset.py: Creates an lmdb dataset to processed/cityscapes/lmdb/.
• models/CVPPP, models/cityscapes: Stores checkpoints of the trained models.
• outputs/CVPPP, outputs/cityscapes: Stores predictions of the trained models.

Data Preparation

Data should be prepared prior to training and evaluation.

• Activate previously created conda environment : source activate ins-seg-pytorch

CVPPP

• Place the extracted dataset to instance-segmentation-pytorch/data/raw/CVPPP/. Hence, raw dataset should be found at instance-segmentation-pytorch/data/raw/CVPPP/CVPPP2017_LSC_training/.
• In order to prepare the data go to instance-segmentation-pytorch/data/CVPP/scripts and
  • python 1-create_annotations.py
  • sh 1-remove_alpha.sh
  • python 2-get_image_paths.py
  • python 3-create_dataset.py

Cityscapes

• Place the extracted datasets to instance-segmentation-pytorch/data/raw/cityscapes/. Hence, raw dataset should be found at instance-segmentation-pytorch/data/raw/cityscapes/gtFine/ and instance-segmentation-pytorch/data/raw/cityscapes/leftImg8bit/.
• In order to prepare the data go to instance-segmentation-pytorch/data/cityscapes/scripts and
  • python 1-create_annotations.py
  • python 2-get_image_paths.py
  • python 3-create_dataset.py

Visdom Server

Start a Visdom server in a screen or tmux.

• Activate previously created conda environment : source activate ins-seg-pytorch

• Start visdom server : python -m visdom.server

• Access visdom server using http://localhost:8097

Training

• Activate previously created conda environment : source activate ins-seg-pytorch

• Go to instance-segmentation-pytorch/code/ and run train.py.

usage: train.py [-h] [--model MODEL] [--usegpu] [--nepochs NEPOCHS]
                [--batchsize BATCHSIZE] [--debug] [--nworkers NWORKERS]
                --dataset DATASET

optional arguments:
  -h, --help            show this help message and exit
  --model MODEL         Filepath of trained model (to continue training)
                        [Default: '']
  --usegpu              Enables cuda to train on gpu [Default: False]
  --nepochs NEPOCHS     Number of epochs to train for [Default: 600]
  --batchsize BATCHSIZE
                        Batch size [Default: 2]
  --debug               Activates debug mode [Default: False]
  --nworkers NWORKERS   Number of workers for data loading (0 to do it using
                        main process) [Default : 2]
  --dataset DATASET     Name of the dataset: "cityscapes" or "CVPPP"

Debug mode plots pixel embeddings to visdom, it reduces size of the embeddings to two-dimensions using TSNE. Hence, it slows down training.

As training continues, models will be saved to instance-segmentation-pytorch/models/CVPPP or instance-segmentation-pytorch/models/cityscapes according to the dataset argument.

Evaluation

After training is complete, we can make predictions.

• Activate previously created conda environment : source activate ins-seg-pytorch

• Go to instance-segmentation-pytorch/code/.

• Run pred_list.py.

usage: pred_list.py [-h] --lst LST --model MODEL [--usegpu]
                    [--n_workers N_WORKERS] --dataset DATASET

optional arguments:
  -h, --help            show this help message and exit
  --lst LST             Text file that contains image paths
  --model MODEL         Path of the model
  --usegpu              Enables cuda to predict on gpu
  --n_workers N_WORKERS
                        Number of workers for clustering
  --dataset DATASET     Name of the dataset: "cityscapes" or "CVPPP"

For example: python pred_list.py --lst ../data/metadata/CVPPP/validation_image_paths.txt --model ../models/CVPPP/2018-3-4_16-15_jcmaxwell_29-937494/model_155_0.123682662845.pth --usegpu --n_workers 4 --dataset CVPPP

• After prediction is completed we can run evaluate.py. It will print metrics to the stdout.

usage: evaluate.py [-h] --pred_dir PRED_DIR --dataset DATASET

optional arguments:
  -h, --help           show this help message and exit
  --pred_dir PRED_DIR  Prediction directory
  --dataset DATASET    Name of the dataset: "cityscapes" or "CVPPP"

For example: python evaluate.py --pred_dir ../outputs/CVPPP/2018-3-4_16-15_jcmaxwell_29-937494-model_155_0.123682662845/validation/ --dataset CVPPP

Prediction

After training is complete, we can make predictions. We can use pred.py to make predictions for a single image.

• Activate previously created conda environment : source activate ins-seg-pytorch

• Go to instance-segmentation-pytorch/code/.

• Run pred.py.

usage: pred.py [-h] --image IMAGE --model MODEL [--usegpu] --output OUTPUT
               [--n_workers N_WORKERS] --dataset DATASET

optional arguments:
  -h, --help            show this help message and exit
  --image IMAGE         Path of the image
  --model MODEL         Path of the model
  --usegpu              Enables cuda to predict on gpu
  --output OUTPUT       Path of the output directory
  --n_workers N_WORKERS
                        Number of workers for clustering
  --dataset DATASET     Name of the dataset: "cityscapes" or "CVPPP"

Results

CVPPP

Scores on validation subset (28 images)

SBD	|DiC|	Foreground Dice
87.4	0.6	96.8

Sample Predictions

Cityscapes

Scores on validation subset

Sample Predictions

TODO

• PEP-8 Style coding.
• Support batch predictions.

Cityscapes

• Train a model.
• Update evaluation script to support cityscapes.
• Improve data creation scripts.
• Add results.