Deep-Learning-Based Embedded ADAS System
ResumoADAS (Advanced Driver Assistance System) applications are computational systems which help drivers in decision making on day-to-day traffic situations. We propose an ADAS application using deep-learning techniques applied in computer vision with the objective to lane segmentation and object detection of common subjects in traffic environment. All the fore mentioned detections are applied to alert a driver about possible collisions with cars, or nearby people. Furthermore, the lane detection has the objective of infer if the driver is deviating from the correct track path. We also propose that the application is deployed and usable in embedded platforms such as NVIDIA’s Jetson family products, to measure applicability we use FPS as a metric both on Jetson Nano and Jetson TX2.
Palavras-chave: Deep learning, Power demand, Roads, Computational modeling, Graphics processing units, Object detection, Libraries, component, formatting, style, styling, insert
SOUSA, Frederico Luiz Martins de; SILVA, Maurício José da; SANTOS, Ricardo Creonte Câmara de Meira; SILVA, Mateus Coelho; OLIVEIRA, Ricardo Augusto Rabelo. Deep-Learning-Based Embedded ADAS System. In: SIMPÓSIO BRASILEIRO DE ENGENHARIA DE SISTEMAS COMPUTACIONAIS (SBESC), 11. , 2021, Evento Online. Anais [...]. Porto Alegre: Sociedade Brasileira de Computação, 2021 . p. 111-118. ISSN 2237-5430.