Reliable segmentation of thermal facial images in unconstrained settings such as thermal ambience and occlusions is challenging as facial features lack salience. Limited availability of datasets from such settings further makes it difficult to train segmentation networks. To address the challenge, we propose Self-Adversarial Multi-scale Contrastive Learning (SAM-CL) as a generic learning framework to train segmentation networks. SAM-CL framework constitutes SAM-CL loss function and a thermal image augmentation (TiAug) as a domain-specific augmentation technique to simulate unconstrained settings based upon existing datasets collected from controlled settings. We use the Thermal-Face-Database to demonstrate effectiveness of our approach. Experiments conducted on the existing segmentation networks- UNET, Attention-UNET, DeepLabV3 and HRNetv2 evidence the consistent performance gain from the SAM-CL framework. Further, we present a qualitative analysis with UBComfort and DeepBreath datasets to discuss how our proposed methods perform in handling unconstrained situations.

This work was published at British Machine Vision Conference, 2022.

While this repository provides the main SAM-CL code for training and validation of thermal datasets, the inference only code along with pre-trained model is available on this repo.

• Linux (Ubuntu ≥ 18.04) with Python ≥ 3.6
• PyTorch ≥ 0.4.1
• torchvision that matches the PyTorch installation. You can install them together at pytorch.org to make sure of this.
• You may use pip install -r requirements.txt to install the dependencies.

We thank the authors of the Thermal Face Project for providing the data. Thermal Face Database can be requested from the authors of Thermal Face Project. From the data that is obtained, please note the path of "FaceDB_Snapshot_complete" directory.

We arrange images and labels in a specific way. You could preprocess the files by running following command with the paths in square-bracket, specified as per the paths on your local PC:

python lib/datasets/preprocess/thermalFaceDB/thermalFaceDB_generator.py --save_dir [~/dev/data/ThermalFaceDB_LM/Processed] --ori_root_dir [~/dev/data/ThermalFaceDB_Org/FaceDB_Snapshot_complete]

and finally, the dataset directory should look like below under the DATA_ROOT directory:

$ThermalFaceDBx320
├── Processed
│   ├── train
│   │   ├── image
│   │   └── label
│   ├── val
│   │   ├── image
│   │   └── label

Note: This implementation is built on openseg.pytorch as well as ContrastiveSeg. We would like to thank the developers of both these repositories for providing the source code.

Replace all the fields within square brackets with actual path and appropriate experiment name (e.g. --config [~/dev/data/ThermalFaceDB])

bash [scripts/thermalFaceDB/deeplab/run_x_8_deeplabv3_train_samcl_occ.sh] train [x_8_samcl_occ] [~/dev/data/ThermalFaceDB] [~/dev/data/ThermalFaceDB]

Replace all the fields within square brackets with actual path and appropriate experiment name (e.g. --config [~/dev/data/ThermalFaceDB])

bash [scripts/thermalFaceDB/deeplab/run_x_8_deeplabv3_train_samcl_occ.sh] val [x_8_samcl_occ] [~/dev/data/ThermalFaceDB] [~/dev/data/ThermalFaceDB]

@inproceedings{Joshi_2022_BMVC,
author    = {Jitesh N Joshi and Nadia Berthouze and Youngjun Cho},
title     = {Self-adversarial Multi-scale Contrastive Learning for Semantic Segmentation of Thermal Facial Images},
booktitle = {33rd British Machine Vision Conference 2022, {BMVC} 2022, London, UK, November 21-24, 2022},
publisher = {{BMVA} Press},
year      = {2022},
url       = {https://bmvc2022.mpi-inf.mpg.de/0864.pdf}
}