The object recognition task using Neural Network requires lots of computing power in order to provide high throughput. Traditional embedded system with RISC processor cannot cope with the demands. However, with the assistance of the hardware accelerator running on the FPGA, the throughput on such a system can be significantly increased.