Real-time 3D Object Detection using Feature Map Flow

TL;DR

This paper introduces a real-time 3D object detection method that leverages temporal feature map aggregation to enhance detection accuracy while maintaining high speed, validated on nuScenes and Waymo datasets.

Contribution

The paper proposes the feature map flow (FMF) technique, a novel method for aggregating temporal features to improve 3D detection performance in real-time.

Findings

Improved detection accuracy over baseline methods.

Achieved real-time performance on nuScenes and Waymo benchmarks.

Code implementation is publicly available.

Abstract

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