Presentation
LightFusion: Lightweight CNN Architecture for Enabling Efficient Sensor Fusion in Autonomous Driving
DescriptionWe propose a framework, named LightFusion, to develop one lightweight and accurate CNN architecture (LA-RoadNet) for enabling efficient fusion. Firstly, our LA-RoadNet is constructed by maintaining the same number of fusion stages in the baseline model and further respectively cutting down original amount of basic blocks in each fusion stage of this baseline model, thus greatly reducing the computational overhead. Then, for gaining high perception accuracy, we introduce a joint unbalanced loss for guiding our LA-RoadNet to both mimic the baseline model's structured information and learn from original ground-truth labels together. Our LA-RoadNet gains 4.8x speedup compared with baseline model.
Event Type
Work-in-Progress Poster
TimeTuesday, July 11th6:00pm - 7:00pm PDT
LocationLevel 2 Lobby
AI
Autonomous Systems
Cloud
Design
EDA
Embedded Systems
RISC-V
Security
