Diffusion-Based Particle-DETR for BEV Perception

TL;DR

This paper introduces a diffusion-based DETR model for BEV perception in autonomous vehicles, improving small object detection and uncertainty modeling by combining diffusion paradigms with 3D detectors.

Contribution

It proposes a novel diffusion-based DETR model with object query interpolation, enhancing small object detection and uncertainty modeling in BEV perception.

Findings

Achieves equal or better performance than state-of-the-art deterministic methods on NuScenes.

Effectively detects small objects in large BEV coverage.

Demonstrates the effectiveness of diffusion-based uncertainty modeling.

Abstract

The Bird-Eye-View (BEV) is one of the most widely-used scene representations for visual perception in Autonomous Vehicles (AVs) due to its well suited compatibility to downstream tasks. For the enhanced safety of AVs, modeling perception uncertainty in BEV is crucial. Recent diffusion-based methods offer a promising approach to uncertainty modeling for visual perception but fail to effectively detect small objects in the large coverage of the BEV. Such degradation of performance can be attributed primarily to the specific network architectures and the matching strategy used when training. Here, we address this problem by combining the diffusion paradigm with current state-of-the-art 3D object detectors in BEV. We analyze the unique challenges of this approach, which do not exist with deterministic detectors, and present a simple technique based on object query interpolation that allows the model to learn positional dependencies even in the presence of the diffusion noise. Based on this, we present a diffusion-based DETR model for object detection that bears similarities to particle methods. Abundant experimentation on the NuScenes dataset shows equal or better performance for our generative approach, compared to deterministic state-of-the-art methods. Our source code will be made publicly available.