IMKD: Intensity-Aware Multi-Level Knowledge Distillation for Camera-Radar Fusion

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Mishra, Shashank, Patil, Karan, Stricker, Didier, Rambach, Jason
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909967718547456
author Mishra, Shashank
Patil, Karan
Stricker, Didier
Rambach, Jason
author_facet Mishra, Shashank
Patil, Karan
Stricker, Didier
Rambach, Jason
contents High-performance Radar-Camera 3D object detection can be achieved by leveraging knowledge distillation without using LiDAR at inference time. However, existing distillation methods typically transfer modality-specific features directly to each sensor, which can distort their unique characteristics and degrade their individual strengths. To address this, we introduce IMKD, a radar-camera fusion framework based on multi-level knowledge distillation that preserves each sensor's intrinsic characteristics while amplifying their complementary strengths. IMKD applies a three-stage, intensity-aware distillation strategy to enrich the fused representation across the architecture: (1) LiDAR-to-Radar intensity-aware feature distillation to enhance radar representations with fine-grained structural cues, (2) LiDAR-to-Fused feature intensity-guided distillation to selectively highlight useful geometry and depth information at the fusion level, fostering complementarity between the modalities rather than forcing them to align, and (3) Camera-Radar intensity-guided fusion mechanism that facilitates effective feature alignment and calibration. Extensive experiments on the nuScenes benchmark show that IMKD reaches 67.0% NDS and 61.0% mAP, outperforming all prior distillation-based radar-camera fusion methods. Our code and models are available at https://github.com/dfki-av/IMKD/.
format Preprint
id arxiv_https___arxiv_org_abs_2512_15581
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle IMKD: Intensity-Aware Multi-Level Knowledge Distillation for Camera-Radar Fusion
Mishra, Shashank
Patil, Karan
Stricker, Didier
Rambach, Jason
Computer Vision and Pattern Recognition
Machine Learning
68T45
I.2.10; I.4.8; I.5.1
High-performance Radar-Camera 3D object detection can be achieved by leveraging knowledge distillation without using LiDAR at inference time. However, existing distillation methods typically transfer modality-specific features directly to each sensor, which can distort their unique characteristics and degrade their individual strengths. To address this, we introduce IMKD, a radar-camera fusion framework based on multi-level knowledge distillation that preserves each sensor's intrinsic characteristics while amplifying their complementary strengths. IMKD applies a three-stage, intensity-aware distillation strategy to enrich the fused representation across the architecture: (1) LiDAR-to-Radar intensity-aware feature distillation to enhance radar representations with fine-grained structural cues, (2) LiDAR-to-Fused feature intensity-guided distillation to selectively highlight useful geometry and depth information at the fusion level, fostering complementarity between the modalities rather than forcing them to align, and (3) Camera-Radar intensity-guided fusion mechanism that facilitates effective feature alignment and calibration. Extensive experiments on the nuScenes benchmark show that IMKD reaches 67.0% NDS and 61.0% mAP, outperforming all prior distillation-based radar-camera fusion methods. Our code and models are available at https://github.com/dfki-av/IMKD/.
title IMKD: Intensity-Aware Multi-Level Knowledge Distillation for Camera-Radar Fusion
topic Computer Vision and Pattern Recognition
Machine Learning
68T45
I.2.10; I.4.8; I.5.1
url https://arxiv.org/abs/2512.15581