Road Sign Classification with Denoising Pipeline Approach

Yu Letian, Sun Yuqi, Zhang Jiaqi

email of author

---

This repository is the PyTorch implementation of our idea: Road Sign Classification with Denoising Pipeline Approach, which the paper is accepted by ICoIAS

Contents

1. Structure
2. Dataset
3. Training
4. Testing
5. Results
6. Citation
7. License and Acknowledgement

Structure

Dataset

# Directory structure of the dataset
GTSRB/
  ├── train/
  │    ├── 00000/
  │    │    ├── 00000_00000.ppm
  │    │    ├── 00000_00001.ppm
  │    │    ├── ...
  │    │    └── GT-00000.csv
  │    ├── 00001/
  │    ├── ...
  │    ├── 00042/
  │    └── Readme-Images.txt
  ├── test/
  │    ├── Images/
  │    │    ├── 00000.ppm
  │    │    ├── 00001.ppm
  │    │    └── ...
  │    └── GT-final_test.csv

Training

To train the model, run the following command:

  # to train a resnet18 model with clean images
  # 1. modify in load.py, change root_dir to 'GTSRB/train'
  # 2. run the training loop below, select the corresponding file
  python train_resnet_dirty.py

Testing

To test the single model, run the following command:

  # to test a resnet18 model with clean images
  # 1. modify in load.py, change root_dir to 'GTSRB/test/Images'
  # 2. modify in load.py, change img_dir='GTSRB/test',csv_file to 'GTSRB/test/GT-final_test.csv'
  python evaluate.py

To test the pipeline approach, run the following command:

  # to test a resnet18 model with denoised images
  # 1. modify in load.py, change root_dir to 'GTSRB/test/Images'
  # 2. modify in load.py, change img_dir='GTSRB/test',csv_file to 'GTSRB/test/GT-final_test.csv'
  python evaluate_pipeline.py

License-and-Acknowledgement

This repository is licensed under the Apache License 2.0.

It includes code from SwinIR, which is also licensed under the Apache License 2.0. Please also follow their licenses.

It includes code from LoRA, which is licensed under the MIT License and adopts the Microsoft Open Source Code of Conduct. Please also follow their licenses.

It contains code for training and testing models on the GTSRB dataset. (J. Stallkamp, M. Schlipsing, J. Salmen, and C. Igel. The German Traffic Sign Recognition Benchmark: A multi-class classification competition. In Proceedings of the IEEE International Joint Conference on Neural Networks, pages 1453–1460. 2011.)

Thanks for their awesome works.