The field of 3D reconstruction is witnessing significant advancements with the integration of multimodal approaches, combining visual and depth information to improve reconstruction quality. Researchers are exploring the limits of traditional methods, investigating the necessity of image guidance in single-view image guided point cloud completion and developing innovative frameworks that can operate without it. Diffusion-based models are also being applied to depth completion tasks, demonstrating improved generalization capabilities and robustness in real-world scenarios. Furthermore, self-supervised learning pipelines are being developed to tackle the challenge of 3D reconstruction from partially occluded images. Noteworthy papers in this area include:

• A Strong View-Free Baseline Approach for Single-View Image Guided Point Cloud Completion, which proposes a strong baseline approach for SVIPC based on an attention-based multi-branch encoder-decoder network.
• DidSee: Diffusion-Based Depth Completion for Material-Agnostic Robotic Perception and Manipulation, which achieves state-of-the-art performance on multiple benchmarks and demonstrates robust real-world generalization.
• DeOcc-1-to-3: 3D De-Occlusion from a Single Image via Self-Supervised Multi-View Diffusion, which introduces an end-to-end framework for occlusion-aware multi-view generation and provides a standardized protocol for evaluating future methods under partial occlusions.