Smart Embedded System for Skin Cancer Classification

The very good results achieved with recent algorithms for image classification based on deep learning have enabled new applications in many domains. medical field is one that can greatly benefit from these order to help the professional elaborate his/her diagnostic. In particular, portable devices are useful scenarios where a full analysis system not an option or difficult obtain. Algorithms models computationally demanding; therefore, it run them low-cost low energy consumption and high efficiency. this paper, proposed classify skin cancer images. Two approaches were followed achieve fast accurate system. At algorithmic level, cascade inference technique was considered, two used inference. architectural processing unit Vitis-AI considered design efficient accelerators FPGA. dual model trained implemented detection ZYNQ UltraScale+ MPSoC ZCU104 evaluation kit ZU7EV device. core integrated system-on-chip solution tested HAM10000 dataset. It achieves performance of 13.5 FPS accuracy 87%, only 33k LUTs, 80 DSPs, 70 BRAMs 1 URAM.