OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Mask3D

Mask3D is a Python research model and implementation that uses a mask-transformer approach to predict semantic instances in 3D point clouds. Its codebase includes model components, dataset preprocessing, Hydra configuration and training code, and reports results on ScanNet, ScanNet200, S3DIS and STPLS3D.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Mask3D — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Mask3D when a local research implementation for 3D point-cloud instance segmentation fits your dataset and you can handle its environment setup and data preprocessing.

Who it is for

  • Researchers studying 3D computer vision and point-cloud instance segmentation.
  • Teams benchmarking semantic-instance segmentation on ScanNet, ScanNet200, S3DIS or STPLS3D.
  • Developers able to work with a local Python and CUDA research environment.

Documented uses

  • Predict semantic instances in 3D point-cloud scenes.
  • Prepare supported benchmark datasets for the model workflow.
  • Explore or extend a modular segmentation research codebase.
  • Run training and benchmarking workflows using the repository’s training code and scripts.
Before you choose

Requirements & limits

The implementation is available in a public GitHub repository under the MIT license.

  • The documented environment specifies Python 3.10.9 and CUDA 11.3.
  • The setup instructions include PyTorch, PyTorch Lightning, Hydra, MinkowskiEngine, Detectron2 and other dependencies.
  • ScanNet and ScanNet200 require dataset preprocessing, including test-scene segmentations from ScanNet’s Segmentator.
  • The README says S3DIS requires a manual dataset fix before preprocessing.
  • The project reports results on four named datasets, but the supplied material does not establish suitability for other datasets.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Prepare ScanNet or ScanNet200 data

Preprocess ScanNet-family data for the Mask3D workflow.

Before you start

  • A Mask3D environment with the documented dependencies installed.
  • Raw ScanNet data and a ScanNet Git repository.
  • Test-scene segmentations generated with the ScanNet Segmentator and placed in ./data/raw/scannet_test_segments.

Steps

  1. Run the ScanNet Segmentator on test scenes using its default parameters.

  2. Save the generated segmentations in ./data/raw/scannet_test_segments.

  3. Run the ScanNet preprocessing module with the raw dataset path, output path, ScanNet repository path, and the appropriate scannet200 setting.

Expected outcome

The preprocessing command is configured to write processed data under the selected output directory, such as data/processed/scannet.

Guide limitations

  • The guide covers preprocessing, not model training or inference.
  • ScanNet200 uses the same preprocessing entry point with a different scannet200 setting.
  • The README does not provide a verified result for a particular machine or dataset.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
JonasSchult/Mask3D · Official homepage title · retained in its source language
How we review sourcesSuggest a correction