Intelligent skin detection system based on multimodal depth learning

Skin detection technology is an important branch in the field of image recognition, and has been widely used in the field of computer vision. It has great potential and good application prospects.Malignant melanoma is a kind of skin cancer with high mortality rate. Its incidence rate is increasing year by year, and has the characteristics of fast metastasis and poor prognosis. Early diagnosis and clinical intervention are the most effective methods for prognosis and mortality reduction. However, because the number of doctors with melanoma detection ability is difficult to match the incidence rate, and the detection and diagnosis of tumor by naked eye are subjective, the reproducibility of the diagnosis is not ideal. At present, melanoma detection technology based on dermoscopic image is an effective way to solve these problems, but the complexity of dermoscopic image itself, such as blurred lesion boundary, intra group variation and inter group similarity, poses a great technical challenge to melanoma detection. Based on these problems, this paper designs a deep learning technology using convolutional neural network, which can automatically learn hierarchical features from data, avoid complex manual extraction, and show advanced results in the field of natural image and medical image analysis.The system adopts multi-modal deep learning. The image information collected by the image sensor is combined with metadata such as user age and skin past medical history for multi-modal feature fusion and collaborative learning. Then, the detection results of freckle and melanoma in the skin are obtained through full convolution neural network and deep convolution neural network. In this paper, a region segmentation model of melanoma lesions based on full convolution neural network is proposed, and a series of effective training strategies are proposed accordingly.It has great use value and research value