3M3D: Multi-view, Multi-path, Multi-representation for 3D Object Detection

TL;DR

The paper introduces 3M3D, a novel 3D object detection method that updates multi-view and query features through multi-attention mechanisms, significantly improving performance on autonomous driving benchmarks.

Contribution

It proposes a multi-view, multi-path, multi-representation framework that enhances scene understanding by updating features with self-attention and multi-representation queries.

Findings

Improves 3D detection accuracy on nuScenes dataset.

Enhances global and local scene understanding through multi-view feature updates.

Achieves performance gains over baseline models.

Abstract

3D visual perception tasks based on multi-camera images are essential for autonomous driving systems. Latest work in this field performs 3D object detection by leveraging multi-view images as an input and iteratively enhancing object queries (object proposals) by cross-attending multi-view features. However, individual backbone features are not updated with multi-view features and it stays as a mere collection of the output of the single-image backbone network. Therefore we propose 3M3D: A Multi-view, Multi-path, Multi-representation for 3D Object Detection where we update both multi-view features and query features to enhance the representation of the scene in both fine panoramic view and coarse global view. Firstly, we update multi-view features by multi-view axis self-attention. It will incorporate panoramic information in the multi-view features and enhance understanding of the global scene. Secondly, we update multi-view features by self-attention of the ROI (Region of Interest) windows which encodes local finer details in the features. It will help exchange the information not only along the multi-view axis but also along the other spatial dimension. Lastly, we leverage the fact of multi-representation of queries in different domains to further boost the performance. Here we use sparse floating queries along with dense BEV (Bird's Eye View) queries, which are later post-processed to filter duplicate detections. Moreover, we show performance improvements on nuScenes benchmark dataset on top of our baselines.