This repository contains the code and models related to the research paper "Optimizing Pothole Detection in Pavements: A Comparative Analysis of Deep Learning Models" by Tiago Tamagusko and Adelino Ferreira. This paper was presented at the Second International Conference on Maintenance and Rehabilitation of Constructed Infrastructure Facilities in Honolulu, HI, USA, 16–19 August 2023.
The paper delves into the exploration of state-of-the-art computer vision techniques for detecting pavement potholes, comparing the performance of several deep learning models based on the You Only Look Once (YOLO) family. The models were trained and tested on a dataset containing 665 road pavement images with labeled potholes. The findings revealed that YOLOv4 yielded the highest mean average precision (mAP), while YOLOv4-tiny offered the optimal reduced inference time, making it suitable for mobile applications. Additionally, the YOLOv5s model demonstrated potential by showcasing impressive results and ease of implementation and scalability.
-
Clone the repository to your local machine.
git clone https://github.com/tamagusko/pothole-detection.git
-
Create a Python virtual environment and activate it:
python -m venv my_env_name source my_env_name/bin/activate
-
Install PyTorch:
pip install torch==1.10.2+cu113 torchvision==0.11.3+cu113 torchaudio==0.10.2+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html
-
Clone YOLOv5:
git clone https://github.com/ultralytics/yolov5.git
-
Install the requirements:
cd yolov5 pip install -r requirements.txt
To train the model:
python path/to/yolov5/train.py --img 300 --batch 16 --epochs 300 --data path/to/data/data_local.yaml --weights yolov5s.pt --project 'project_name' --name 'run_name'
-
Pull the docker image:
sudo docker pull ultralytics/yolov5:latest
-
Run the docker image:
sudo docker run -v path/to/data/:/mnt/data --ipc=host -it --network host --gpus all ultralytics/yolov5:latest
-
Train the model:
python train.py --img 300 --batch 16 --epochs 300 --data /mnt/data/data.yaml --weights yolov5s.pt --project "test" --name "run1"
If you find this repository useful for your research, please cite our paper as follows:
@inproceedings{tamagusko2023optimizing,
title={Optimizing Pothole Detection in Pavements: A Comparative Analysis of Deep Learning Models},
author={Tamagusko, Tiago and Ferreira, Adelino},
booktitle={Second International Conference on Maintenance and Rehabilitation of Constructed Infrastructure Facilities},
year={2023},
location={Honolulu, HI, USA}
}
For issues, bug reports, and pull requests, please use this GitHub page. To contact me directly, send an email to tamagusko@gmail.com.
This work is licensed under a CC-BY-NC-ND-4.0 license. (c) 2023, Tiago Tamagusko.