RADE-Net: Robust Attention Network for Radar-Only Object Detection in Adverse Weather
Authors: Christof Leitgeb, Thomas Puchleitner, Max Peter Ronecker, Daniel Watzenig
Organizations: 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 · Institute of Visual Computing, Graz University of Technology, Graz, Austria
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.
Figures & tables
Fig. 1: Proposed architecture: (I) Tensor Projection Module (TPM), (II) Encoder-Decoder Backbone, (III) Dilated Neck and (IV) Range-Azimuth based Centerpoint Detection Heads + Cartesian Transformation.
Fig. 2: Tensor Projection Module (TPM): The full FFT-processed Radar tensor is split into Range-Azimuth-Doppler (RAD) and Range-Azimuth-Elevation (RAE) projections by calculating the maximum along each excluded dimension. The projections are subsequently concatenated along the third dimension.
Fig. 3: RADE-Net backbone with three down-sampling stages and CBAM supported skip connections.
Modality
Metric
Networks
Total
Normal
Overcast
Fog
Rain
Sleet
Light Snow
Heavy Snow
Lidar
AP3D
Voxel-RCNN [ 44 ]
46.4
81.8
69.6
48.8
47.1
46.9
54.8
37.2
CasA [ 45 ]
50.9
82.2
65.6
44.4
53.7
49.9
62.7
36.9
TED-S [ 46 ]
51.0
74.3
68.8
45.7
53.6
44.8
63.4
36.7
Camera +Lidar
AP3D
VPFNet [ 47 ]
52.2
81.2
76.3
46.3
53.7
44.9
63.1
36.9
TED-M [ 46 ]
52.3
77.2
69.7
47.4
54.3
45.2
64.3
36.3
MixedFusion [ 48 ]
55.1
* 84.5*
* 76.6*
53.3
55.3
49.6
* 68.7*
44.9
TABLE I: Sedan detection comparison under various weather conditions
Object class
Average Precision (AP)
Total
Normal
Fog
Rain
Sleet
Light Snow
Heavy Snow
3D
BEV
3D
BEV
3D
BEV
3D
BEV
3D
BEV
3D
BEV
3D
BEV
Sedan
56.75
63.21
55.10
61.29
83.09
89.13
49.11
58.69
39.96
44.45
62.10
72.22
61.82
63.33
Bus or Truck
40.99
48.29
37.25
40.77
-
-
1.05
1.05
14.12
17.99
68.96
72.56
65.94
85.94
Pedestrian
5.40
6.79
0.99
0.99
15.43
19.94
0.99
0.99
1.40
1.60
-
-
-
-
Bicycle
1.49
1.49
1.49
1.49
-
-
-
-
-
-
-
-
-
-
TABLE II: Evaluation of our RADE-Net with multiple road users and under various weather conditions
Feature
Input
Dilated
Expanded
CBAM
Base
IoU=0.3
IoU=0.5
Expansion
Stem
Neck
Heads
AP3D
APBEV
AP3D
APBEV
✓
57.96
63.98
32.85
56.18
✓
✓
55.13
62.88
27.65
50.99
✓
✓
✓
58.35
64.70
32.73
55.98
✓
✓
✓
59.81
66.32
35.31
57.01
✓
✓
✓
✓
64.06
68.71
39.02
61.64
TABLE III: Ablation Study
Fig. 4: Model performance visualization under different weather conditions. The ground truth bounding boxes are shown in white along with class label and the model predictions are shown in red along with class label and confidence score. The ROI where detections are considered is marked with a dashed black rectangle.
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.
Christof Leitgeb, Thomas Puchleitner, Max Peter Ronecker +1
Infineon Technologies AG, Austria · Graz University of Technology, Austria · Virtual Vehicle Research GmbH, Austria
Perception is a fundamental component of autonomous driving systems. While LiDAR-based methods have achieved remarkable progress in object detection, their reliability can degrade under adverse weather conditions. Radar point clouds provide a robust alternative due to their resilience to bad weather and low-illumination scenarios. However, radar point clouds are typically sparse, unordered, and less informative than LiDAR data, making it challenging to directly apply existing LiDAR-based perception methods. To address these challenges, we propose IRGNN, an Invariant Radar Graph Neural Network for radar point cloud object detection. IRGNN first reconstructs radar point clouds into graph representations using translation- and rotation-invariant feature designs, enabling robust modeling of sparse radar measurements. It then employs an improved message passing neural network (MPNN) with residual connections and a virtual node layer to enhance local feature propagation and global context modeling. Finally, task-specific heads are applied to the learned graph representations for object classification and bounding box prediction. Experimental results on the RadarScenes dataset show that IRGNN outperforms existing radar-based object detection methods and achieves competitive performance. In addition, IRGNN significantly reduces computational cost and memory usage during inference, demonstrating its effectiveness and practical potential for efficient radar-based perception in autonomous driving.
Xiao Guo, Wanke Xia, Lili Yang +1
China Agricultural University · Tsinghua University
Robust 3D object detection in adverse weather conditions is challenging due to sensor limitations. Although combining complementary modalities such as LiDAR and 4D RADAR has shown promise, the sparsity of these sensors becomes apparent in adverse weather with reduced reflections, leading to objects with few or no point cloud returns. To address this limitation, camera sensors provide visual cues even when LiDAR and RADAR signals are weakened. However, cameras themselves are also vulnerable to adverse weather, where some regions become unreliable due to snow or rain occluding the camera lens. While some camera-fusion methods designed for adverse weather learn to weigh image regions via confidence maps, these maps receive no direct supervision and are learned solely through the detection loss. We introduce Reliability-Aware Fusion (RAF), which explicitly supervises per-pixel reliability estimation and provides a direct learning signal for identifying and suppressing unreliable visual cues. Our framework leverages pretrained LiDAR-RADAR networks, keeping their backbones frozen while only training the added camera branch, BEV fusion encoder, and detection head. Extensive experiments on the K-Radar and VoD datasets demonstrate that integrating RAF consistently improves detection accuracy over LiDAR-RADAR baselines, achieving up to +6.5 APBEV and +7.4 AP3D gains. Code is available at https://github.com/parkie0517/RAF.