The project titled “Hardware-Software Integrated and Quantized Deep Learning Accelerator Design on FPGA for ADAS Systems,” led by Asst. Prof. Dr. Güner TATAR, has been awarded support within the scope of the TÜBİTAK 1005 - National New Ideas and Products Research Support Program. In addition to our faculty member, the research team includes Prof. Dr. Burcu ERKMEN from Yıldız Technical University, Assoc. Prof. Dr. Salih BAYAR from Marmara University, and Asst. Prof. Dr. İhsan ÇİÇEK from Gebze Technical University as researchers.
The project aims to develop an energy-efficient and reconfigurable FPGA-based ADAS solution as an alternative to high power-consuming GPU-based solutions. By doing so, it intends to increase Türkiye’s competitiveness in automotive and smart transportation technologies.
This project aims to develop a hardware-software integrated, real-time, and energy-efficient deep learning accelerator based on FPGA for Advanced Driver Assistance Systems (ADAS). Within the scope of the project, critical ADAS tasks such as object detection, lane detection, drivable area segmentation, depth estimation, and semantic segmentation will be executed concurrently on a single hardware architecture using quantized (INT8/INT16) deep learning models. The proposed system is based on a hardware–software co-design approach integrating an ARMbased processing system with FPGA programmable logic. By employing quantization-aware training and hardware-level optimizations, the project aims to achieve significantly lower power consumption and latency compared to conventional GPU-based solutions. The expected outcomes of the project include contributions to domestic ADAS technologies, support for university-industry collaboration, and the creation of commercialization opportunities in the automotive, defense, and intelligent transportation sectors.
We congratulate our faculty member and wish him continued success.
