Embedded Neural Networks for Medical Image Classification

Embedded Neural Networks for Medical Image Classification

Nanda Kumar E L1, M Sai Charan Teja2, Manoj K3, Manoj S4, Dr. Kiran Bailey5

[1][2][3][4] UG Students, DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING, BMS College of Engineering (Affiliated to VTU), Bangalore, India.

[5] Assistant Professor, DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING, BMS College of Engineering (Affiliated to VTU), Bangalore, India.

---------------------------------------------------------------------***---------------------------------------------------------------------

Abstract - This paper “Embedded Neural Networks for Medical Image Classification” aims to improvise speed of execution of Neural networks by using specialized hardware architecture for computation in FPGA. With recent advancements in medical technology, the biomedical field has ushered in the era of big data. The applications of deep learning in medical image analysis, electronic health record, genomics and drug development has been found. Deep learning has obvious advantages in making full use of biomedical data and improving medical health level. Thus, there is a need for a hardware framework for inference in embedded medical applications which use deep learning. Deep learning is computationally intensive, real time embedded systems which use deep learning require fast computing hardware to meet its timing requirements. To meet these requirements, the project utilizes and surveys FINN framework which is a compiler framework for generating data flow accelerators for Neural Network models.

Key Words:  Image classification, FINN, FPGA, Quantized Neural Networks, Inference engine.