3DMV

Third Workshop for Learning 3D with Multi-View Supervision @ CVPR 2026

The third 3DMV workshop at CVPR 2026 explores the rapidly expanding frontier of 3D and 4D research from multi-view data. Building on the success of the first and second editions that focused on 3D and object generation, 3DMV 2026 highlights dynamic multi-view datasets, generative 4D models, and the emerging paradigm of world models—systems that learn to predict and simulate evolving 3D scenes with spatial and temporal consistency. We also explore their integration with frontier techniques such as LLMs and video diffusion. A special emphasis is placed on emerging applications in robotics and medical AI, where 3D/4D world models enable robust generalization. 3DMV 2026 aims to unify 2D, 3D, and 4D research communities and foster emerging applications toward scalable, cross-domain multi-view learning. The detailed topics covered in the workshop include the following:

Multi-View for 3D Understanding

Deep Multi-View Stereo

Multi-View for 3D Generation and Novel View Synthesis

Dynamic Multi-View Datasets and 4D Generative models

LLMs and World Models for 3D/4D

Video Diffusion for Multi-View Generation

Robotics Applications with Multi-View 3D/4D

Medical AI with Multi-View 3D/4D

Call For Papers

We are soliciting papers that use Multi-view deep learning to address problems in 3D/4D Understanding and Generation, including but not limited to the following topics:

Bird-Eye View for 3D Object Detection

Multi-view fusion for 3D Object Detection

Indoor/outdoor scenes segmentation

3D/4D Diffusions for generation

Video diffusion for multi-view synthesis

4D understanding, generation, and world models

LLMs and foundation models for 3D/4D

Language + 3D/4D

Medical 3D segmentation and analysis

Robotics with multi-view 3D/4D perception

3D shape generation and reconstruction

Deep multi-view stereo

Inverse Graphics from multi-view images

Indoor/outdoor scenes generation and reconstruction

Volumetric Multi-view representation for 3D generation and novel view synthesis

NeRFs and Gaussian Splatting

3D shape classification and retrieval

Vision for XR, AR, VR

Paper Submission Guidelines

We accept both archival and non-archival paper submissions.

Archival submissions should be of max 8 pages (excluding references) on the aforementioned and related topics.

Non-archival submissions can be previously published works in major venues (in the last two years or at CVPR 2026) or based on new works (max 8 pages as well).

Archival Accepted papers will be included in the proceedings of CVPR 2026, while non-archival accpeted papers will not be included

Submitted manuscripts should follow the CVPR 2026 paper template (if they have not been published previously).

All submissions (except for previoulsly published) will be peer-reviewed under a double-blind policy (authors should not include names in submissions)

PDFs need to be submitted online through the link.

Accepted papers' authors will be notified to prepare camera-ready posters to be uploaded based on the schedule above.

Every accepted paper will have the opportunity to host a poster presentation at the workshop.

Some acccpeted papers will be selcted for oral presentations at the workshop.

There will be a `best poster award` announced during the workshop with a sponsored money prize.