Presentation
R-TOSS: A Framework for Real-Time Object Detection using Semi-Structured Pruning
DescriptionObject detectors used in autonomous vehicles can have high memory and computational overheads. In this paper, we introduce a novel pruning framework that overcomes shortcomings of state-of-the-art model compression techniques. Experiment results on the JetsonTX2 shows that our framework has a compression rate of 4.4x on YOLOv5 with a 2.15x speed up in inference time and 57.01% decrease in energy usage from that of the base model. The approach also allows 2.89x compression on RetinaNet with a 1.86x speedup in inference time and 56.31% decrease in energy usage while outperforming the original model in terms of accuracy.
Event Type
Research Manuscript
TimeThursday, July 13th1:55pm - 2:10pm PDT
Location3003, 3rd Floor
Autonomous Systems
Cloud
Autonomous Systems (Automotive, Robotics, Drones)