Skip to content
VirtualOnlinePast event

Best of 3DV 2026: Livestream Research Showcase

Event artwork for Best of 3DV 2026: Livestream Research Showcase

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

This online event offers researchers, engineers, and computer vision practitioners a curated virtual pass to the most significant research from the 3D Vision Conference (3DV) 2026. With 177 accepted papers spanning 3.5 days, attending every session at a major conference is impossible — this livestream series solves that challenge by bringing the best work directly to you.

What You'll Experience

Join leading researchers as they present four cutting-edge papers that defined this year's conference. Each session features live presentations from the original authors, followed by opportunities to engage with the research community.

Featured Research

Navigating Conference Content with VLMs and Embeddings — Harpreet Sahota, machine learning engineer and hacker-in-residence, demonstrates how to build an interactive map of an entire conference using open-source tools. This practical session shows how document embeddings and vision language models can transform how we explore research literature, using the complete 3DV 2026 corpus as a case study.

Seeing Through Clutter: Structured 3D Scene Reconstruction — Rio Aguina-Kang from Drafted AI presents a novel approach to reconstructing 3D scenes from single images. Rather than relying on traditional semantic segmentation, this method uses an iterative object removal pipeline orchestrated by VLMs, achieving state-of-the-art robustness on 3D-Front and ADE20K datasets.

Physical Realistic 4D Generation — Lu Sang, PhD researcher at TU Munich and Google XR intern, explores the frontier of dynamic 3D content. Learn about TwoSquared (3DV 2026 oral presentation), which achieves full 4D generation from just two 2D image pairs using physically constrained velocity fields.

Finding NeMO: Few-Shot Object Perception — Sebastian Jung from the German Aerospace Center presents NeMO, a geometry-aware representation enabling 6DoF object pose estimation, detection, and segmentation from just a handful of RGB images — no model retraining required.

Who Should Attend

This event is ideal for machine learning engineers, computer vision researchers, robotics developers, and anyone working with 3D reconstruction, visual computing, or spatial AI. Whether you're building AR/VR applications, working on physical AI systems,

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