Deep learning in diabetic foot ulcers detection: A comprehensive evaluation

There has been a substantial amount of research involving computer methods and technology for the detection recognition diabetic foot ulcers (DFUs), but there is lack systematic comparisons state-of-the-art deep learning object frameworks applied to this problem. DFUC2020 provided participants with comprehensive dataset consisting 2,000 images training testing. This paper summarises results by comparing learning-based algorithms proposed winning teams: Faster R-CNN, three variants R-CNN an ensemble method; YOLOv3; YOLOv5; EfficientDet; new Cascade Attention Network. For each method, we provide detailed description model architecture, parameter settings additional stages including pre-processing, data augmentation post-processing. We evaluation method. All required stage increase number available post-processing remove false positives. The best performance was obtained from Deformable Convolution, variant mean average precision (mAP) 0.6940 F1-Score 0.7434. Finally, demonstrate that method based on different can enhanced not mAP.