In this work, a simple yet effective deep neural network is proposed to generate the dense depth map of the scene by exploiting both LiDAR sparse point cloud and the monocular camera image. Specifically, a feature pyramid network is firstly employed to extract feature maps from images across time. Then the relative pose is calculated by minimizing the feature distance between aligned pixels from inter-frame feature maps. Finally, the feature maps and the relative pose are further applied to compute the feature-metric loss for training the depth completion network. The key novelty of this work lies in that a self-supervised mechanism is presented to train the depth completion network by directly using visual-LiDAR odometry between consecutive frames. Comprehensive experiments and ablation studies on benchmark dataset KITTI demonstrate the superior performance over other state-of-the-art methods in terms of pose estimation and depth completion. The detailed performance of the proposed approach (referred to as SelfCompDVLO ) can be found on the KITTI depth completion benchmark. The source code, models, and data have been made available at GitHub.

In the realm of category-level 6D object pose estimation, canonical 3D representation reconstruction is pivotal, yet current methods show limitations in reconstruction quality, a key step in current pose estimation pipeline. To address this, we introduce an innovative Adversarial Canonical Representation Reconstruction Network (ACR-Pose) in this paper. In particular, ACR-Pose comprises a Reconstructor, with novel sub-modules: a Pose-Irrelevant Module (PIM) for robustness to rotation and translation, and a Relational Reconstruction Module (RRM) for extracting relational information between input modalities. A Discriminator is incorporated to guide the generation of realistic canonical representations through adversarial optimization. Evaluated on the prevalent NOCS-CAMERA and NOCS-REAL datasets, our method significantly improves the performance of baseline models and achieves comparable performance with existing state-of-the-art methods, representing a promising advancement in the field of category-level 6D object pose estimation.

3D reconstruction of urban scenes is an important research topic in remote sensing. Neural Radiance Fields (NeRFs) offer an efficient solution for both structure recovery and novel view synthesis. The realistic 3D urban models generated by NeRFs have potential future applications in simulation for autonomous driving, as well as in Augmented and Virtual Reality (AR/VR) experiences. Previous NeRF methods struggle with large-scale, urban environments. Due to the limited model capability of NeRF, directly applying them to urban environments may result in noticeable artifacts in synthesized images and inferior visual fidelity. To address this challenge, we propose a sparse voxel-based NeRF. First, our approach leverages LiDAR odometry to refine frame-by-frame LiDAR point cloud alignment and derive accurate initial camera pose through joint LiDAR-camera calibration. Second, we partition the space into sparse voxels and perform voxel interpolation based on 3D LiDAR point clouds, and then construct a voxel octree structure to disregard empty voxels during subsequent ray sampling in the NeRF, which can increase the rendering speed. Finally, the depth information provided by the 3D point cloud on each viewpoint image supervises our NeRF model, which is further optimized using a depth consistency loss function and a plane constraint loss function. In the real-world urban scenes, our method significantly reduces the training time to around an hour and enhances reconstruction quality with a PSNR improvement of 1–2 dB, outperforming other state-of-the-art NeRF models.