DehazeDET: Deep Image Dehazing for Improving Object Detection

Overview

DehazeDET is a novel dual-branch architecture that integrates the DSEU model (https://github.com/yeanwei97/DSEU) for image restoration and the Mix-YOLONet detection network for object detection into a joint parallel framework.

Features

• Training: Train the model using train.py.
• Testing: Test the model's performance using train.py --test.
• Inference: Perform inference on a single image or an entire folder of images.
• Data Conversion: Convert various datasets (VOC, Foggy Driving, RTTS) to the required format.

Dataset

• VOC-FOG: Download this model at https://drive.google.com/file/d/1bLUtwrKwzPwLI3yZBFZYw4BnINpxCfVp/view?usp=sharing (from https://github.com/yz-wang/TogetherNet) .
• RTTS: Download this model at https://sites.google.com/view/reside-dehaze-datasets/reside-%CE%B2.
• Foggy Driving: Download this model at https://people.ee.ethz.ch/~csakarid/SFSU_synthetic/.

Usage

Training

To train the model, run:

python train.py

Testing

To test the model, run:

python train.py --test --data your_dataset

Inference

To perform inference on a single image, run:

python train.py --inference --image_path path_to_your_image

To perform inference on an entire folder of images, run:

python train.py --inference_test --model_path your_model --inference_input your_input_image

Data Conversion

Convert VOCFOG dataset:

python voctoyolo.py 

Convert Foggy Driving dataset:

python foggytovoc.py 

Convert RTTS dataset:

python rttstovoc.py 

Configuration

Modify the args.yaml file to change training parameters and other settings.

License

This project is licensed under the MIT License - see the LICENSE file for details

Acknowledgements

• DSEU by yeanwei97(https://github.com/yeanwei97/DSEU)
• Mix-YOLONet