Skip to content
VirtualOnlinePast event

May 11 - Best of 3DV 2026

Event artwork for May 11 - Best of 3DV 2026

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

Experience cutting-edge 3D computer vision research from the comfort of your home with this virtual showcase of standout papers from the 3DV 2026 conference. This online event, organized by the San Diego AI, ML and Computer Vision Meetup, streams directly from the authors themselves, offering rare insight into some of the most impactful work in 3D scene understanding, 4D generation, and few-shot object perception.

Why This Event Stands Out

With 177 papers accepted to 3DV 2026, navigating a major computer vision conference can feel overwhelming. This curated Best of 3DV series solves that problem by bringing together a handpicked selection of innovative research talks, letting you learn from the researchers behind breakthrough papers without the travel costs or scheduling conflicts of in-person attendance.

Featured Talks and Speakers

Navigating a 3D Vision Conference with VLMs and Embeddings — Harpreet Sahota, machine learning engineer and hacker-in-residence, demonstrates how to build an interactive conference map using modern open-source tools. Using document embeddings from Jina v4 and Qwen3.5-9B for annotation, he shows how to process all 177 3DV papers into a queryable FiftyOne dataset with UMAP visualization, similarity search, and representativeness scoring. This practical approach to conference navigation uses VLMs for topic classification, author affiliation extraction, and project page detection — perfect for anyone overwhelmed by large academic conferences.

Seeing Through Clutter: Structured 3D Scene Reconstruction — Rio Aguina-Kang from Drafted AI presents a novel method for reconstructing 3D representations from single images by iteratively removing and modeling objects one at a time. Using VLMs as orchestrators, this approach handles complex occluded scenes better than traditional semantic segmentation methods. With previous experience at Adobe Research, Brown Visual Computing, and Stanford's HAI Institute, Aguina-Kang demonstrates state-of-the-art robustness on 3D-Front and ADE20K datasets — all without task-specific training.

Physical Realistic 4D Generation — Lu Sang, PhD researcher at TU Munich working with Prof. Daniel Cremers, takes you through a series of works on generating dynamic 3D content that moves and deforms over time. From neural surface deformation with explicit velocity fields (ICLR 2025) to the 4Deform framework (CVPR 2025) and TwoSquared (3DV 2026 oral), which achieves full 4D generation from just tw

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