DVMN: Dense Validity Mask Network for Depth Completion

Reichardt, Laurenz; Mangat, Patrick; Wasenmüller, Oliver

doi:10.1109/itsc48978.2021.9564588

Cited by 5 publications

(2 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Sensor fusion is a method that has proven to counteract the sparsity of LiDAR data [16] and is a natural approach to complement radars' missing height information. Image data are dense and include height information but lack depth information, which radar can provide.…”

Section: Three-dimensional Radar Perceptionmentioning

confidence: 99%

“…Sparse data can pose a challenge for traditional convolution layers as initially, most of the information in the filter's receptive field is empty. With progressive layers, the valid information is propagated through the network and increases in density [16,29,32]. For this reason, while early low-level features can represent small structures due to their large spatial size, the sparsity of this information can be complemented by high-level features with denser information.…”

Section: Sparsity-robust Feature Fusionmentioning

confidence: 99%

See 1 more Smart Citation

Sparsity-Robust Feature Fusion for Vulnerable Road-User Detection with 4D Radar

Ruddat,

Reichardt,

Ebert

et al. 2024

Applied Sciences

Self Cite

View full text Add to dashboard Cite

Detecting vulnerable road users is a major challenge for autonomous vehicles due to their small size. Various sensor modalities have been investigated, including mono or stereo cameras and 3D LiDAR sensors, which are limited by environmental conditions and hardware costs. Radar sensors are a low-cost and robust option, with high-resolution 4D radar sensors being suitable for advanced detection tasks. However, they involve challenges such as few and irregularly distributed measurement points and disturbing artifacts. Learning-based approaches utilizing pillar-based networks show potential in overcoming these challenges. However, the severe sparsity of radar data makes detecting small objects with only a few points difficult. We extend a pillar network with our novel Sparsity-Robust Feature Fusion (SRFF) neck, which combines high- and low-level multi-resolution features through a lightweight attention mechanism. While low-level features aid in better localization, high-level features allow for better classification. As sparse input data are propagated through a network, the increasing effective receptive field leads to feature maps of different sparsities. The combination of features with different sparsities improves the robustness of the network for classes with few points.

show abstract

Section: Three-dimensional Radar Perceptionmentioning

confidence: 99%

Section: Sparsity-robust Feature Fusionmentioning

confidence: 99%