This research investigates the performance of four unified convolutional neural networks (CNNs) based on the You Only Look Once (YOLO) model in recognizing skin cancer lesions. The study focuses on three types of lesions: malignant melanoma, benign nevi, and seborrheic keratosis. The models evaluated are YOLOv3, YOLOv4, YOLOv5, and YOLOv7, with YOLOv7 demonstrating superior performance. The experiments were conducted using the International Skin Imaging Collaboration (ISIC) dataset, which contains 2750 dermoscopy images. The performance metrics used include Intersection over Union (IoU), Mean Average Precision (mAP), F1-measure, and inference time. YOLOv7 achieved an IoU of 86.3%, mAP of 75.4%, F1-measure of 80%, and an inference time of 0.32 seconds per image. The study highlights the potential of YOLOv7 as a valuable tool for aiding dermatologists in early skin cancer diagnosis and reducing unnecessary biopsies. The research also discusses the importance of data augmentation to address class imbalances and improve model performance.This research investigates the performance of four unified convolutional neural networks (CNNs) based on the You Only Look Once (YOLO) model in recognizing skin cancer lesions. The study focuses on three types of lesions: malignant melanoma, benign nevi, and seborrheic keratosis. The models evaluated are YOLOv3, YOLOv4, YOLOv5, and YOLOv7, with YOLOv7 demonstrating superior performance. The experiments were conducted using the International Skin Imaging Collaboration (ISIC) dataset, which contains 2750 dermoscopy images. The performance metrics used include Intersection over Union (IoU), Mean Average Precision (mAP), F1-measure, and inference time. YOLOv7 achieved an IoU of 86.3%, mAP of 75.4%, F1-measure of 80%, and an inference time of 0.32 seconds per image. The study highlights the potential of YOLOv7 as a valuable tool for aiding dermatologists in early skin cancer diagnosis and reducing unnecessary biopsies. The research also discusses the importance of data augmentation to address class imbalances and improve model performance.