A Novel Image Processing System to Recognize and Classify Images using Integrated Advanced CNN and FPGA for IoT based Applications

FPGAs are ideal for deep convolutional neural network (CNN) acceleration because of their high parallelism and energy economy. Automatic compilers for the generation of network-specific FPGA accelerators have been developed in recent years. The use of network-specific accelerators, in combination with more cascade deep CNN algorithms, necessitates running-time re-configuration of FPGA devices. Edge devices may have a hard time with this kind of reconfiguration. FPGAs are a potential option for embedded applications. There are several uses for the Zynq-7000 plus FPGA, including a dual-core ARM Cortex-A9 processing system with double wrap up. The problem statement prompted the development of a hybrid model combination of ACNN and FPGA. FPGA provides rich analogous computing resources with high power competence, making it ideal for deep convolutional neural network (CNN) acceleration. Employing CNN with FPGA is a stimulating work. FPGA executions ensure memory restrictions and fall short to encounter source necessities for critical tasks. Proposed system yields good detection rate i.e 0.08 and improved frame processing rate as 124.82. Here detection time also observed to be reduced as 9sec. furthermore proposed one outperforms to improve efficiency levels.