Design And Development Of Deep Learning Convolutional Neural Network On An Field Programmable Gate Array

Lee, Yan Qing (2018) Design And Development Of Deep Learning Convolutional Neural Network On An Field Programmable Gate Array. Project Report. UTeM, Melaka, Malaysia. (Submitted)

Text (24 Pages) Design And Development Of Deep Learning Convolutional Neural Network On An Field Programmable Gate Array.pdf - Submitted Version Download (237kB)

Abstract

This work presents design and development of Deep Learning Convolutional Neural Network (CNN) on an Field Programmable Gate Array (FPGA). In recent, CNN is a challenging research area in terms both software and hardware. However, software implementations tend to be prohibitively slow due to more of the neural networks run on sequentially operation architecture. The objective of this work is to develop the deep learning CNN on FPGA since hardware implementations perform parallel computation of each neuron in the layers can be made faster. FPGA are construction of programmable logic, which are not only erasable and flexible for design the realize the algorithm like the software, but also have a great speed to operate some kinds of algorithm due to FPGA has parallel execution ability. This work focuses on handwriting recognition where the machines has the ability to receive and interpret intelligible handwritten input from the sources. Researchers all over the world have achieved successful results in handwritten recognition which can be divided into handwritten numeral recognition, character and cursive word recognition. Neural network is the way people used to realize the pattern classification and image recognition. The design in this work utilize filter which acts as feature detectors from the original input image to extracted and recognized the patterns in images. The speed and the accuracy of the CNN implemented on an FPGA are analysed. Digits and numbers are successfully recognized by the system.

Actions (login required)

View Item