Search papers, labs, and topics across Lattice.
This study evaluates Spiking Neural Networks (SNNs) for automotive multi-object detection and tracking, addressing the high energy consumption of traditional deep learning methods. By employing transfer learning with the SpikeYOLO architecture, the authors achieved competitive performance metrics on established datasets, including a mean Average Precision of 0.937 on KITTI and 0.771 on BDD100K MOT2020 for object detection. The findings highlight SNNs as a viable, energy-efficient alternative for perception tasks in autonomous systems, potentially transforming deployment strategies in edge computing environments.
SNNs achieve competitive automotive detection and tracking performance while significantly reducing energy consumption compared to traditional deep learning models.
Deep learning algorithms are notorious for their high carbon footprint and computational demands that limit their deployment on edge devices and raise concerns about their long-term sustainability. Neuromorphic computing and Spiking Neural Networks (SNNs) offer a promising alternative to traditional Von Neumann architectures, providing energy-efficient performance, massively parallel computation, and on-chip learning capabilities. Autonomous machines represent a critical application domain where these advantages are particularly valuable. We present the first comprehensive evaluation of SNNs for real-world automotive multi-object detection and tracking. Using transfer learning with the SpikeYOLO architecture, we achieve mean Average Precision of 0.937 on the KITTI dataset and 0.771 on BDD100K MOT2020 dataset for object detection and a Higher Order Tracking Accuracy score of 0.701 (KITTI) and 0.445 (BDD100K MOT2020) for object tracking--results competitive with conventional deep learning methods. Our results demonstrate that SNNs can deliver high-performance object detection and tracking in an energy efficient manner, establishing their viability for perception in real-world autonomous systems.
