The client is a US-based automotive division of a Japanese conglomerate, engaged in providing automotive solutions such as car multimedia, ADAS, and autonomous technology. The company had proprietary algorithms for camera calibration and optical self-diagnosis of its ADAS system. However, the code was highly inefficient and needed to be ported from the PC to a suitable platform for car multimedia with high speed and low power consumption.

eInfochips recommended Hydra DSP and executed the porting and optimization of the client’s camera algorithm, tackling with complex technical challenges to come out with an optimal solution.

The key highlights of the project included floating point to fixed point conversion, and conversion of the entire code from C++ to C along with handling complex structures and dynamic memory, thereby overcoming the challenges of restricted memory (32 KB code memory, 4 KB DMem and 4KB DMem 2). The result was accurate, ready-to-deploy ADAS algorithms on the Broadcom chipset.