Skip to content
VirtualOnlinePast event

Best of 3DV 2026: Cutting-Edge 3D Vision Research

Event artwork for Best of 3DV 2026: Cutting-Edge 3D Vision Research

This event has ended

This listing is kept as an archive. Explore upcoming events to find your next one.

See upcoming events
About this event

What to expect

Dive deep into cutting-edge 3D computer vision research with this virtual showcase of the year's most impactful work from the 3DV 2026 conference. With 177 accepted papers spanning just 3.5 days at the original conference, this curated online session brings together leading researchers to present their groundbreaking work directly to a global audience — no travel required.

This online meetup tackles one of the biggest challenges facing conference attendees: how do you absorb meaningful insights from hundreds of papers in a limited time? The answer lies in bringing the best research directly to you through live presentations from the authors themselves.

Featured Research Presentations

Navigating a 3D Vision Conference with VLMs and Embeddings — Learn how to build an interactive map of an entire academic conference using modern open-source tools. Speaker Harpreet Sahota, a hacker-in-residence and machine learning engineer, demonstrates an end-to-end pipeline that processes all 177 papers from 3DV 2026 using VLMs and document embeddings. The system performs topic classification, author affiliation extraction, and project page detection, then creates an explorable dataset for similarity search and uniqueness scoring. Perfect for anyone interested in document AI, VLMs, and practical applications of embedding models.

Seeing Through Clutter: Structured 3D Scene Reconstruction — Rio Aguina-Kang from Drafted AI presents an innovative approach to reconstructing 3D scenes from single images by iteratively removing and modeling objects. This method addresses the limitations of traditional semantic segmentation in complex, occluded scenes by decomposing them into simpler subtasks. The system uses VLMs as orchestrators and achieves state-of-the-art robustness without requiring task-specific training.

Physical Realistic 4D Generation — Lu Sang, PhD researcher at TU Munich, shares breakthrough work on generating dynamic 3D content that moves and deforms over time. The presentation covers neural surface deformation with physically constrained velocity fields and introduces TwoSquared, which achieves full 4D generation from just two 2D image pairs — a practical path toward controllable, physically plausible content creation for VR/AR, robotics, and digital humans.

Finding NeMO: Few-Shot Object Perception — Sebastian Jung from the German Aerospace Center (DLR) presents NeMO, a novel object representation enabling 6DoF object pose estimation, detectio

Running an AI event?

List it here, free.

Meetups, workshops and conferences go in front of thousands of builders after a quick review.

List your event