Mix3D Demo In-the-Wild 3D Scans

Indoor 3D scene understanding has experienced a tremendous growth in the last few years. A major driving force behind these developments are large-scale 3D datasets such as ScanNet, Matterport or S3DIS. Unlike large-scale 2D datasets that are based on in-the-wild image collections uploaded by a large number of internet users, current 3D indoor datasets are recorded by a few experts. This limits the diversity of the depicted scenes and makes it difficult to collect very large datasets which help to train deep models that generalize well.

In-the-Wild 3D Scans

In this demo, we showcase how well current state-of-the-art models for 3D semantic segmentation perform on newly captured "in-the-wild" 3D scans. And you can participate by uploading a 3D scan from anything interesting around you! We will run our approach Mix3D and other recent methods on the uploaded scans and present the semantic segmentation results below.

Join us for our 3DV's Demo Session 1A on Thursday, 2nd December 2021 11:20 (GMT+1) to see Mix3D performing on your very own challenging 3D scans!

Call for Contribution

How do current 3D segmentation approaches like Mix3D or MinkowskiNet perform on 3D scans recorded in the wild? Let's find out! Use your iPhone or iPad equipped with a LiDAR scanner, and upload your 3D scans below. The most interesting scans and their semantic segmentations will be presented below and during our 3DV demo:
Session 1A on Thursday, 2nd December 2021 11:20 (GMT+1).

Upload 3D Scan

Instructions for recording 3D scans using the "3D Scanner App" for iPhone or iPad with a Lidar sensor (3D meshes recorded with other devices are also welcome). For privacy reasons, please do not record people, remove any personal objects from the scene, and only record on locations where you have permission to do so.

Uploaded scans so far: 0


Below, we show 3D semantic segmentation results of Mix3D and MinkowskiNet on "in-the-wild" scenes uploaded by users across the world. The models are trained on the ScanNet dataset.

Cabinet Bed Chair Sofa Table Door Window Bookshelf Picture Counter Desk Curtain
Refrigerator Bathtub Shower curtain Toilet Sink Other furniture


Advising Committee