Resource constrained cellular neural networks for real-time obstacle detection using FPGAs

Due to the fast growing industry of smart cars and autonomous driving, advanced driver assistance systems (ADAS) with its applications have attracted a lot of attention. As a crucial part of ADAS, obstacle detection has been challenge due to the real-tme and resource-constraint requirements. Cellular neural network (CeNN) has been popular for obstacle detection, however suffers from high computation complexity. In this paper we propose a compressed CeNN framework for real-time ADAS obstacle detection in embedded FPGAs. Particularly, parameter quantizaion is adopted. Parameter quantization quantizes the numbers in CeNN templates to powers of two, so that complex and expensive multiplications can be converted to simple and cheap shift operations, which only require a minimum number of registers and LEs. Experimental results on FPGAs show that our approach can significantly improve the resource utilization, and as a direct consequence a speedup up to 7.8x can be achieved with no performance loss compared with the state-of-the-art implementations.