Miftahul Huda1 Dimas Rizky Ramadhani2 Nabila Azhari3 Arsyiah Azahra4 Putri Maulida Chairani5
Detection and classification of litter on the beach is carried out using the Real-Time Detection Transformer (RT-DETR) model. RT-DETR is a transformer-based object detection architecture designed for real-time data processing with high performance https://arxiv.org/abs/2304.08069, making it very suitable for application in environmental monitoring tasks like this.
reference: https://arxiv.org/abs/2304.08069 and https://github.com/ultralytics/ultralytics/tree/main/ultralytics
The plastic litter dataset used is sourced from the following link https://universe.roboflow.com/monash-ventz/beach-waste-vqths with the following data usage license explanation https://creativecommons.org/licenses/by/4.0/. RT-DETR was trained using 2675 training data to recognize several main labels included in the plastic waste category, such as:
- Bottle
- Clothes
- Metal
- Plastic
- Rope
- Styrofoam
- Wood
There are 561 validation data without wood and clothes category for evaluation
| Class | Images | Instances | Precision | Recall | mAP@IoU[50] | mAP@IoU[50-95] |
|---|---|---|---|---|---|---|
| All | 561 | 2558 | 0.848 | 0.744 | 0.810 | 0.606 |
| Bottle | 366 | 874 | 0.906 | 0.814 | 0.872 | 0.638 |
| Metal | 228 | 458 | 0.882 | 0.852 | 0.878 | 0.648 |
| Plastic | 331 | 603 | 0.838 | 0.776 | 0.819 | 0.636 |
| Rope | 155 | 205 | 0.709 | 0.512 | 0.619 | 0.409 |
| Styrofoam | 195 | 418 | 0.904 | 0.765 | 0.861 | 0.697 |
results.mp4
[1] Monash, "Beach Waste Dataset," Roboflow Universe. Roboflow, Oct. 2024. [Online]. Available: https://universe.roboflow.com/monash-ventz/beach-waste-vqths. Accessed: Nov. 19, 2024.
[2] Y. Zhao, W. Lv, S. Xu, J. Wei, G. Wang, Q. Dang, Y. Liu, and J. Chen, "DETRs Beat YOLOs on Real-time Object Detection," arXiv preprint, 2024. [Online]. Available: https://arxiv.org/abs/2304.08069.
[3] Ultralytics, "Ultralytics Neural Network Modules," GitHub repository, 2024. [Online]. Available: https://github.com/ultralytics/ultralytics/tree/main/ultralytics/nn/modules. [Accessed: Dec. 1, 2024].