Better depth from monocular image by reasoning globally and locally with MonoViT MonoViT: Self-Supervised Monocular Depth Estimation with a Vision Transformer arXiv paper abstract https://arxiv.org/abs/2208.03543v1 arXiv PDF paper https://arxiv.org/pdf/2208.03543v1.pdf GitHub https://github.com/zxcqlf/monovit Self-supervised monocular depth estimation is an attractive solution that does not require hard-to-source depth labels for training.

Improve segmentation of occluded objects by using 2 layers to model occlusion with BCNet Occlusion-Aware Instance Segmentation via BiLayer Network Architectures arXiv paper abstract https://arxiv.org/abs/2208.04438v1 arXiv PDF paper https://arxiv.org/pdf/2208.04438v1.pdf GitHub https://github.com/lkeab/BCNet Segmenting highly-overlapping image objects is challenging, because there is typically no distinction between real object contours and occlusion boundaries…

Survey of human action recognition using deep learning Video-based Human Action Recognition using Deep Learning: A Review arXiv paper abstract https://arxiv.org/abs/2208.03775v1 arXiv PDF paper https://arxiv.org/pdf/2208.03775v1.pdf Human action recognition … aim is to accurately describe human actions and their interactions from a previously unseen data sequence acquired by sensors. The ability…

Better matching of partial 3D point clouds by using uncertainty with UTOPIC UTOPIC: Uncertainty-aware Overlap Prediction Network for Partial Point Cloud Registration arXiv paper abstract https://arxiv.org/abs/2208.02712v1 arXiv PDF paper https://arxiv.org/pdf/2208.02712v1.pdf GitHub https://github.com/zhileichen99/utopic High-confidence overlap prediction and accurate correspondences are critical for cutting-edge models to align paired point clouds in a…

Reconstruct 3D unknown object from a single view by mapping image to points with ZeroMesh ZeroMesh: Zero-shot Single-view 3D Mesh Reconstruction arXiv paper abstract https://arxiv.org/abs/2208.02676v1 arXiv PDF paper https://arxiv.org/pdf/2208.02676v1.pdf Single-view 3D object reconstruction is a fundamental and challenging computer vision task that aims at recovering 3D shapes from single-view RGB…

Improve 3D human reconstruction by using occupancy planes with OPlanes Occupancy Planes for Single-view RGB-D Human Reconstruction arXiv paper abstract https://arxiv.org/abs/2208.02817 arXiv PDF paper https://arxiv.org/pdf/2208.02817.pdf Single-view RGB-D human reconstruction with implicit functions is often formulated as per-point classification. … The feature of each 3D location is then used to classify…

Better object detection with few examples by using image rather than proposals with Meta-DETR Meta-DETR: Image-Level Few-Shot Detection with Inter-Class Correlation Exploitation arXiv paper abstract https://arxiv.org/abs/2208.00219v1 arXiv PDF paper https://arxiv.org/pdf/2208.00219v1.pdf GitHub https://github.com/ZhangGongjie/Meta-DETR Few-shot object detection has been extensively investigated by incorporating meta-learning into region-based detection frameworks.

Improve segmentation of close objects in video using optical flow with BATMAN BATMAN: Bilateral Attention Transformer in Motion-Appearance Neighboring Space for Video Object Segmentation arXiv paper abstract https://arxiv.org/abs/2208.01159 arXiv PDF paper https://arxiv.org/pdf/2208.01159.pdf Video Object Segmentation (VOS) is fundamental to video understanding. … existing work faces challenges segmenting visually similar objects…

Complete sparse depth maps from many domains by training on simulated gaps with FillDepth Towards Domain-agnostic Depth Completion arXiv paper abstract https://arxiv.org/abs/2207.14466 arXiv PDF paper https://arxiv.org/pdf/2207.14466.pdf Existing depth completion methods are often targeted at a specific sparse depth type, and generalize poorly across task domains. … present a method to…