Efficient 3D object detection using noise conditioned score network for accurate votes and object proposals.
Enhancing dense BEV frameworks for accurate 3D object detection with BEVNeXt.
3D 객체 감지에서의 Scaling Adversarial Robustness의 중요성
CN-RMA introduces an innovative approach for 3D indoor object detection from multi-view images, achieving state-of-the-art performance by combining reconstruction and detection networks with occlusion-aware feature aggregation.
BEVENet, a novel convolutional-only architecture, achieves state-of-the-art efficiency in 3D object detection for autonomous driving by leveraging Bird's-Eye-View (BEV) space and outperforms computationally intensive Vision Transformer (ViT)-based methods.
ProFusion3D is a novel LiDAR-camera fusion framework for robust 3D object detection that leverages a progressive fusion strategy across different views (BEV and PV) and levels (intermediate features and object queries), enhanced by a self-supervised pre-training method for improved data efficiency.
ROA-BEV improves the accuracy of vision-based 3D object detection in autonomous driving by using 2D region-oriented attention to help the network focus on areas where objects are likely to exist, thereby reducing interference from background information.