The dataset used for this project is the RDD2022 Dataset. It is from 2022 paper RDD2022: A multi-national image dataset for automatic Road Damage Detection.
It consists both drone and car images. In particular, I the RDD2022_China_Drone dataset for training and testing. It is fairly small, with only 2400 images. Nevertheless, it is a good dataset to start with.
| Class Name | Description |
|---|---|
| D00 | Longitudinal Cracks |
| D10 | Transverse Cracks |
| D20 | Alligator Cracks |
| D40 | Potholes |
| Repair | Repaired Crack |
| Block crack | Block Cracks |
| Image 1 | Image 2 |
|---|---|
 |
 |
To start with, I used the YoloV1 model. It is a single stage object detection model. It is simple and easy to implement. It is also fasف. At the time of, I only trained the model for 30 epochs, but it is already able to detect the road damages, albeit not very well.
I also tried to use the YoloV3 model, still in progress...