Surface Defect Detection
Introduction:
β β£ Some defects tend to be miniature and difficult to identify
β β£ Various appearances and characteristics of defects often lead to misclassification of their categories
β β£ Segment and classify defects in the input image by predicting pixel-labelled segmentation masks
β β£ Research and develop deep learning semantic segmentation models for automatic defect detection Objectives:
β β£ Evaluate and compare CNN-based, Transformer-based and hybrid CNN-Transformer models in various aspects
β β£ Explore and implement advanced weakly supervised semantic segmentation algorithms / frameworks
β β£ DatasetsKolektor Surface-Defect Dataset 2 (KolektorSDD2/KSDD2)
β β£ Magnetic Tile Surface Defects
Results:
RANKINGS | |||||
---|---|---|---|---|---|
Fully-Supervised | mIoU | Weakly-Supervised | mIoU | ||
Magnetic Tile | KolektorSDD2 | Magnetic Tile | KolektorSDD2 | ||
83.58 | 79.36 | - | 67.00 | ||
83.19 | 78.92 | - | 39.84 | ||
82.81 | 78.43 | ||||
80.65 | 78.21 | ||||
74.25 | 72.98 | ||||
66.95 | - | ||||
62.64 | - |
Sample Prediction Masks:
Fully-Supervised | Magnetic Tile | ||
---|---|---|---|
KolektorSDD2 | |||
Weakly-Supervised | KolektorSDD2 |