DinoRADE: Full Spectral Radar-Camera Fusion with Vision Foundation Model Features for Multi-class Object Detection in Adverse Weather

Christof Leitgeb; Thomas Puchleitner; Max Peter Ronecker; Daniel Watzenig

cs.CVApr 9, 2026

DinoRADE: Full Spectral Radar-Camera Fusion with Vision Foundation Model Features for Multi-class Object Detection in Adverse Weather

Authors: Christof Leitgeb, Thomas Puchleitner, Max Peter Ronecker, Daniel Watzenig

Organizations: Infineon Technologies AG, Austria · Graz University of Technology, Austria · Virtual Vehicle Research GmbH, Austria

Abstract

Reliable and weather-robust perception systems are essential for safe autonomous driving and typically employ multi-modal sensor configurations to achieve comprehensive environmental awareness. While recent automotive FMCW Radar-based approaches achieved remarkable performance on detection tasks in adverse weather conditions, they exhibited limitations in resolving fine-grained spatial details particularly critical for detecting smaller and vulnerable road users (VRUs). Furthermore, existing research has not adequately addressed VRU detection in adverse weather datasets such as K-Radar. We present DinoRADE, a Radar-centered detection pipeline that processes dense Radar tensors and aggregates vision features around transformed reference points in the camera perspective via deformable cross-attention. Vision features are provided by a DINOv3 Vision Foundation Model. We present a comprehensive performance evaluation on the K-Radar dataset in all weather conditions and are among the first to report detection performance individually for five object classes. Additionally, we compare our method with existing single-class detection approaches and outperform recent Radar-camera approaches by 12.1%. The code is available under https://github.com/chr-is-tof/RADE-Net.

Figures & tables

Figure 1: Overview of the DinoRADE architecture.

Figure 2: Reference points projected from 3D Radar queries to camera image (red) and Radar sensor location (blue).

Figure 3: DinoRADE performance visualization in four different scenarios: (a) university campus, (b) alleyway, (c) highway, and (d) road shoulder. Ground truth bounding boxes and class labels are shown in white. Predicted bounding boxes with predicted class and confidence score are shown in red. The ROI where detections are considered is marked with a dashed black rectangle. Magnified views are indicated in orange.

Methods	Modality	Total	Normal	Overcast	Fog	Rain	Sleet	Light Snow	Heavy Snow
Voxel-RCNN [ 3 ]	L	46.4	81.8	69.6	48.8	47.1	46.9	54.8	37.2
CasA [ 41 ]		50.9	82.2	65.6	44.4	53.7	49.9	62.7	36.9
TED-S [ 42 ]		51.0	74.3	68.8	45.7	53.6	44.8	63.4	36.7
RTNH [ 28 ]	R	47.4	49.9	56.7	52.8	42.0	41.5	50.6	44.5
RTNH+ [ 15 ]		57.6	-	-	-	-	-	-	-
RADE-Net [ 17 ]		66.7	66.3	74.5	82.1	57.2	69.4	63.9	68.9

Table 1: 3D Average Precision (AP) comparison on ’Sedan’ class in all weather conditions using K-Radar v1.1.

Weather	Total		Sedan		Bus or Truck		Pedestrian		Motorcycle		Bicycle
Weather	3D	BEV	3D	BEV	3D	BEV	3D	BEV	3D	BEV	3D	BEV
Total	36.99	39.61	71.38	75.32	54.92	58.93	33.01	38.12	* 3.77	* 3.77	21.89	21.89
Normal	29.04	30.08	71.52	74.13	46.61	49.17	*0.99	*0.99	*3.77	*3.77	22.32	22.32
Overcast	70.89	74.33	74.20	75.16	67.57	73.51	-	-	-	-	-	-
Fog	67.41	75.68	83.84	89.91	-	-	50.98	61.46	-	-	-	-
Rain	24.10	25.97	71.32	76.92	*0.99	*0.99	*0.00	*0.00	-	-	-	-

Table 2: AP and mAP (Total) evaluation of all road users in all weather conditions using K-Radar v2.1. (-) indicates no representation in the test set and (*) indicates a severe under-representation of under 2% of total objects in the train set of the respective weather condition.

Methods	Mod.	Total		Sedan		Bus or Truck		Pedestrian		Motorcycle		Bicycle
Methods	Mod.	3D	BEV	3D	BEV	3D	BEV	3D	BEV	3D	BEV	3D	BEV
RADE-Net [ 17 ]	R	20.93	23.96	56.75	63.21	40.99	48.29	5.40	6.79	0.00	0.00	1.49	1.49
DPFT [ 5 ]	RC	11.65	13.54	38.99	41.02	17.60	20.55	1.66	6.14	0.00	0.00	0.00	0.00
Ours	RC	36.99	39.61	71.38	75.32	54.92	58.93	33.01	38.12	3.77	3.77	21.89	21.89

Table 3: AP and mAP (Total) comparison with Radar-Camera and Radar-only methods for multiple object classes using K-Radar v2.1.

Metric	R	R+C	R+C+W	R+C*+W
$AP_{3D}$	61.65	69.61	71.38	68.43
$AP_{BEV}$	66.68	74.96	75.32	72.42

Table 4: Ablation Study

Figure 4: Examples for partially occluded (1), heavily occluded (2), and fully occluded (3).

Object Detection

Object detection locates and classifies objects in images or video, balancing localization accuracy, robustness, and inference speed.

See all 624 papers

RADE-Net: Robust Attention Network for Radar-Only Object Detection in Adverse Weather
Feb 23, 2026Christof Leitgeb, Thomas Puchleitner, Max Peter Ronecker +13D Object Detection mmWave Radar Sensing
Infineon Technologies AG, Institute of Visual Computing, Graz University of Technology, Graz, Austria · Infineon Technologies AG, Graz, Austria · Virtual Vehicle Research GmbH, Institute of Visual Computing, Graz University of Technology, Graz, Austria +1
RadarXFormer: Robust Object Detection via Cross-Dimension Fusion of 4D Radar Spectra and Images for Autonomous Driving
Mar 16, 2026Yue Sun, Yeqiang Qian, Zhe Wang +3Autonomous Driving Perception 3D Object Detection
Global Institute of Future Technology, Shanghai Jiao Tong University, Shanghai, 200240, China. · School of Automation and Intelligent Sensing, Shanghai Jiao Tong University, Shanghai, 200240, China. · Key Laboratory of System Control and Information Processing, Ministry of Education of China, Shanghai Jiao Tong University, Shanghai, 200240, China. +2
RAF: Reliability-Aware Fusion of Camera, LiDAR, and 4D RADAR for Robust 3D Object Detection in Adverse Weather
Jul 6, 2026Heejun Park, Jaeseok Jeong, Kuk-Jin YoonCamera-Radar Fusion Lidar-Based 3D Object Detection
Visual Intelligence Lab., KAIST

Feb 23, 2026J/K move · Enter open · S save

RADE-Net: Robust Attention Network for Radar-Only Object Detection in Adverse Weather

Christof Leitgeb, Thomas Puchleitner, Max Peter Ronecker, Daniel Watzenig

Infineon Technologies AG, Institute of Visual Computing, Graz University of Technology, Graz, Austria · Infineon Technologies AG, Graz, Austria · Virtual Vehicle Research GmbH, Institute of Visual Computing, Graz University of Technology, Graz, Austria+1

3D Object Detection mmWave Radar Sensing

Open paper AskPDF arXiv

Automotive perception systems are obligated to meet high requirements. While optical sensors such as Camera and Lidar struggle in adverse weather conditions, Radar provides a more robust perception performance, effectively penetrating fog, rain, and snow. Since full Radar tensors have large data sizes and very few datasets provide them, most Radar-based approaches work with sparse point clouds or 2D projections, which can result in information loss. Additionally, deep learning methods show potential to extract richer and more dense features from low level Radar data and therefore significantly increase the perception performance. Therefore, we propose a 3D projection method for fast-Fourier-transformed 4D Range-Azimuth-Doppler-Elevation (RADE) tensors. Our method preserves rich Doppler and Elevation features while reducing the required data size for a single frame by 91.9% compared to a full tensor, thus achieving higher training and inference speed as well as lower model complexity. We introduce RADE-Net, a lightweight model tailored to 3D projections of the RADE tensor. The backbone enables exploitation of low-level and high-level cues of Radar tensors with spatial and channel-attention. The decoupled detection heads predict object center-points directly in the Range-Azimuth domain and regress rotated 3D bounding boxes from rich feature maps in the cartesian scene. We evaluate the model on scenes with multiple different road users and under various weather conditions on the large-scale K-Radar dataset and achieve a 16.7% improvement compared to their baseline, as well as 6.5% improvement over current Radar-only models. Additionally, we outperform several Lidar approaches in scenarios with adverse weather conditions. The code is available under https://github.com/chr-is-tof/RADE-Net.

Abstract

Object Detection

Explore similar work

RADE-Net: Robust Attention Network for Radar-Only Object Detection in Adverse Weather

RadarXFormer: Robust Object Detection via Cross-Dimension Fusion of 4D Radar Spectra and Images for Autonomous Driving

RAF: Reliability-Aware Fusion of Camera, LiDAR, and 4D RADAR for Robust 3D Object Detection in Adverse Weather