XIDLE-Net

This is the implementation of the paper "Inspector gaze-guided multitask learning for explainable structural damage assessment".

Getting Started

Installation

• Clone this repo:

git clone https://github.com/itschenyu/XIDLE-Net.git
cd XIDLE-Net

Dataset

• Please download the dataset from here and then place it in ./XIDLE-Net/dataset.

Pre-trained Weight

• Please download pre-trained weights on ImageNet-22K from here and place it in ./XIDLE-Net/Module/model_data/.

Model Download

• Please download the XIDLE-Net model weight from here and then place it in ./XIDLE-Net/Module/model_weight/.

Training

python RUN.py

Testing

Evaluating the model on the test set:

python TEST.py

Citation

If XIDLE-Net and the eye gaze dataset are helpful to you, please cite them as:

@article{https://doi.org/10.1111/mice.70131,
  author = {Zhang, Chenyu and Liu, Charlotte and Li, Ke and Yin, Zhaozheng and Qin, Ruwen},
  title = {Inspector gaze-guided multitask learning for explainable structural damage assessment},
  journal = {Computer-Aided Civil and Infrastructure Engineering},
  volume = {40},
  number = {30},
  pages = {5824-5841},
  doi = {https://doi.org/10.1111/mice.70131},
  url = {https://onlinelibrary.wiley.com/doi/abs/10.1111/mice.70131},
  eprint = {https://onlinelibrary.wiley.com/doi/pdf/10.1111/mice.70131},
  year = {2025}
}

Note

Part of the codes are referred from MT-UNet project.

The images and damage level labels in the dataset are credited to PEER Hub ImageNet.