Implementation of SSD in PyTorch to Classify Pavement Distress For Video

Please make sure you have Nvidia CUDA installed on your system.

This repository implements SSD (Single Shot MultiBox Detector). This implementation is heavily influenced by the projects high quality, fast, modular reference implementation of ssd in pytorch, this repository aims to classify pavement distress based on SSD using video as an input.

Step-by-step installation

git clone https://github.com/bruhtus/pavement_distress_ssd.git
cd pavement_distress_ssd
# Required packages:
pip install -r requirements.txt

Train

Setting Up Datasets

For COCO dataset, make the folder structure like this:

ssd/data/datasets
|__ annotations
    |_ train.json
    |_ ...
|__ train
    |_ <im-1-name>.jpg
    |_ <im-1-name>.json
    |_ ...
    |_ <im-N-name>.jpg
    |_ <im-N-name>.json
|__ ...

Use labelme to do labeling stuff and then use labelme2coco.py to generate COCO data formatted JSON.

Test

Please see documentation.md for more detailed usage of the testing implementation.

Develop Guide

If you want to add your custom components, please see DEVELOP GUIDE on lufficc repo for more details.

Citations

If you use this project in your research, please cite this project.

@misc{bruhtus2020,
    author = {Robertus Diawan Chris},
    title = {{Implementation of SSD in PyTorch to Classify Pavement Distress on Video},
    year = {2020},
    howpublished = {\url{https://github.com/bruhtus/pavement_distress_ssd}}
}