This paper has been accepted at ICIP 2024. This repository provides the benchmark code and checkpoints to evaluate our CO2Wounds-V2 dataset, which contains 764 RGB images of chronic wounds acquired from 96 leprosy patients, with wound semantic segmentation annotations provided in COCO and image formats.
Download the CO2Wounds-V2 dataset free, easy, and fast here.
Clone our repo to your local machine using the following command:
Prerequisites
Create a new conda environment using the provided environment.yml file.
| Architectures | Encoder | mIoU(%) | F1(%) | Accuracy(%) | Precision(%) | Recall(%) | Checkpoints |
|---|---|---|---|---|---|---|---|
| DeepLabV3 | ResNeXt-50 | 68.48 | 77.86 | 98.58 | 85.81 | 78.20 | Download |
| DeepLabV3+ | ResNeXt-50 | 68.23 | 78.04 | 98.51 | 81.61 | 81.69 | Download |
| U-Net | ResNeXt-50 | 69.94 | 79.44 | 98.65 | 84.35 | 80.31 | Download |
| FPN | ResNeXt-50 | 68.99 | 78.36 | 98.53 | 82.94 | 81.17 | Download |
| DeepLabV3+ | ResNet-101 | 66.88 | 76.55 | 98.40 | 83.04 | 79.02 | Download |
| U-Net | ResNet-101 | 66.96 | 76.61 | 98.28 | 80.87 | 81.10 | Download |
| FPN | ResNet-101 | 66.81 | 76.52 | 98.45 | 81.78 | 79.61 | Download |
| DeepLabV3+ | EfficientNet | 66.98 | 76.78 | 98.49 | 79.88 | 81.86 | Download |
| U-Net | EfficientNet | 67.71 | 77.20 | 98.51 | 83.29 | 77.80 | Download |
| FPN | EfficientNet | 67.49 | 76.84 | 98.59 | 82.63 | 80.34 | Download |
| U-Net | SegFormer | 70.13 | 79.26 | 98.59 | 84.70 | 81.99 | Download |
| FPN | SegFormer | 69.90 | 79.36 | 98.56 | 82.02 | 84.35 | Download |
If you use this dataset/code in your research, please cite:
@article{sanchez2024co2wounds,
title={CO2Wounds-V2: Extended Chronic Wounds Dataset From Leprosy Patients},
author={Sanchez, Karen and Hinojosa, Carlos and Mieles, Olinto and Zhao, Chen and Ghanem, Bernard and Arguello, Henry},
journal={arXiv preprint arXiv:2408.10827},
year={2024}
}
The authors make data publicly available according to open data standards and license datasets under the Creative Commons Attribution-NonCommercial-NoDerivatives (CC BY-NC-ND) license.
- Karen Sanchez
Linkedin: https://www.linkedin.com/in/karenyanethsanchez/ Twitter: @karensanchez119 Email: karen.sanchez@kaust.edu.sa