FDADNet: Detection of Surface Defects in Wood-Based Panels Based on Frequency Domain Transformation and Adaptive Dynamic Downsampling

⭐ This code has been completely released ⭐

⭐ our article ⭐

If our code is helpful to you, please cite:

@article{li2024fdadnet,
  title={FDADNet: Detection of Surface Defects in Wood-Based Panels Based on Frequency Domain Transformation and Adaptive Dynamic Downsampling},
  author={Li, Hongli and Yi, Zhiqi and Wang, Zhibin and Wang, Ying and Ge, Liang and Cao, Wei and Mei, Liye and Yang, Wei and Sun, Qin},
  journal={Processes},
  volume={12},
  number={10},
  pages={2134},
  year={2024},
  publisher={Multidisciplinary Digital Publishing Institute}
}

Requirements

pip install -r requirements.txt

Train

1. Prepare training data

• The download link for the WBP-DET data set is here.
• The download link for the NEU-DET data set is here.
• The download link for the GC10-DET data set is here.
• The download link for the APDDD data set is here.

FDADNet
├── WBP-DET
│   ├── images
│   │   ├── 1.jpg
│   │   ├── 2.jpg
│   │   ├── .....
│   ├── labels
│   │   ├── 1.txt
│   │   ├── 2.txt
│   │   ├── .....

2. Begin to train

python train.py

Test

1. Weight

• The download link for the weight is here.

2. Begin to test

python val.py

Results

Methods	OS	GS	Sc	Ch	OD	mAP50	Params/M $\downarrow$	GFLOPs $\downarrow$
Faster R-CNN	64.9	65.7	65.5	96	75.8	73.6	41.37	134
YoloV5s	57.3	56.1	62.2	98	68.3	68.4	7.02	15.8
YoloX-Tiny	67.9	65.8	45.8	93.9	60.5	66.8	5.03	7.57
RTMDet	52.9	69.9	86.1	97.6	82.2	77.8	4.87	8.02
YoloV7-Tiny	61.2	74.6	40.5	97.2	70.1	68.7	6.01	13.1
YoloV8n	57.9	61.9	59.9	97.4	79	71.2	3.01	8.1
YoloV10n	51.1	64.8	63.1	97.3	68.9	69.1	2.7	8.2
RTDETR	56	56.7	77.9	91.8	74.6	71.4	31.9	103.5
FDADNet	61.3	79.3	85.5	98.4	73.7	79.6	4.5	6.2

• Bold indicates first or second best performance.

Time

2024.8.3 Upload code

2024.8.3 Upload requirements.txt

2024.8.6 Upload Dataset

2024.10.14 Upload readme

Visualization of results

Acknowledgements

This code is built on ultralytics (PyTorch). We thank the authors for sharing the codes.

In the comparative experiment, Faster RCNN, YOLOX, and RTMDet networks were replicated using mmdetection. We thank the authors for sharing the codes.

Contact

If you have any questions, please contact me by email ([email protected]).