Real-time 3D Object Detection using Feature Map Flow
This work addresses the need for efficient and accurate 3D detection in autonomous driving, representing an incremental improvement over existing center-based baselines.
The paper tackles real-time 3D object detection by introducing a feature map flow (FMF) method that aggregates time-spatial features from different time steps, improving detection quality and achieving real-time performance on nuScenes and Waymo benchmarks.
In this paper, we present a real-time 3D detection approach considering time-spatial feature map aggregation from different time steps of deep neural model inference (named feature map flow, FMF). Proposed approach improves the quality of 3D detection center-based baseline and provides real-time performance on the nuScenes and Waymo benchmark. Code is available at https://github.com/YoushaaMurhij/FMFNet