Core Concepts
EffiPerception is a versatile framework that enhances accuracy-speed performance with low memory cost across various perception tasks.
Abstract
EffiPerception addresses the accuracy-speed-memory trade-off in computer vision perception tasks.
The framework consists of three main parts: Efficient Feature Extractors, Efficient Layers, and EffiOptim.
Extensive experiments on KITTI, Semantic-KITTI, and COCO datasets demonstrate the effectiveness of EffiPerception.
EffiPerception shows improvements in accuracy, speed, and memory usage compared to existing methods.
Stats
Previous methods focus on single or specific tasks.
EffiPerception achieves great accuracy-speed performance with low memory cost.
Quotes
"EffiPerception could show great accuracy-speed-memory overall performance increase within the four detection and segmentation tasks."
"We proposed an efficient framework called EffiPerception to achieve better accuracy-speed performance with relatively low memory cost under several perception tasks."