| IP개요 |
This project proposes an accuracy configurable energy-efficient approximate convolutional neural network (CNN) processor for mobile applications. The conventional CNN-based systems have been suffered from power hungry operation which caused by floating point operation and in-efficient memory accesses. Solving huge power consumption, the proposed fixed-point arithmetic operation composed with micro memory operation. The bit precision-based operation can provide reasonable energy efficiency with the proper accuracy for optimized mobile systems. As mentioned above, the bit precision-based method will also control accuracy in real-time processing. The proposed CNN processor can achieve state-of-the-art energy efficient technique with the scaled accuracy. All the functionalities of the proposed work are verified by software simulations using Pytorch and verified by RTL simulation using Xilinx FPGA. Also, we will synthesize the proposed hardware with power-aware P&R. As a result, the proposed system is fabricated in 65nm CMOS process, this system could be realization for state-of-the art NN based processor hardware design. |
