Melanoma is the most common skin cancer, and it is increasing widely. Automatic skin lesion detection from dermoscopic images remains a challenging task. Many efforts have been dedicated to this challenge using various methods, but due to its poor robustness, it is not good for the analysis of melanoma skin lesions. Propose a method for skin lesion detection and classification tasks simultaneously to make sure feature learning is successful. The base of feature pyramid networks and region proposal networks is ResNet50, which is used here. The network learns features more quickly using a three-phase cooperative training technique. Before entering this model, the hairs from the images are removed
Copyrights © 2024