Defective sewing stitch semantic segmentation using DeeplabV3+ and EfficientNet

Defective stitch inspection is an essential part of garment manufacturing quality assurance. Traditional mechanical defect detection systems are effective, but they usually customized with handcrafted features that must be operated by a human. Deep learning approaches have recently demonstrated exceptional performance in wide range computer vision applications. The requirement for precise detail evaluation, combined the small size patterns, undoubtedly increases difficulty identification. Therefore, image segmentation (semantic segmentation) was employed this task. It identified as vital research topic field vision, being indispensable real-world Semantic method labeling each pixel image. This direct contrast to classification, which assigns single label entire And multiple objects same class defined entity. DeepLabV3+ architecture, encoder-decoder proposed technique. EfficientNet models (B0-B2) were applied encoders experimental processes. encoder utilized encode feature maps from input encoder's significant information used decoder upsampling and reconstruction output. Finally, best model DeeplabV3+ EfficientNetB1 can classify segmented defective sewing stitches superior (MeanIoU: 94.14%).