Open3D-ML: 3D Machine Learning Framework for LiDAR and RGB-D Data

Open3D-ML is a repository by isl-org updated through February 2026 that provides standardized access to 3D machine learning datasets and pretrained models. It integrates with PyTorch and TensorFlow to facilitate tasks like semantic segmentation and 3D object detection using LiDAR and RGB-D data.

Use Cases

• Training semantic segmentation models using the SemanticKITTI or S3DIS wrappers
• Evaluating 3D object detection performance on LiDAR point clouds with pretrained models
• Visualizing RGB-D data and 3D perception results using the Open3D engine

Strengths

• Supports both PyTorch and TensorFlow frameworks for 3D tasks
• Includes pretrained models for 3D object detection and semantic segmentation
• Provides standardized wrappers for major 3D datasets like SemanticKITTI and S3DIS
• Integrated visualization tools for RGB-D and LiDAR data

Limitations

• Does not host the raw data directly; users must download datasets from original sources
• Metadata and label quality vary by the specific underlying dataset used
• High computational requirements for processing large-scale 3D point clouds

Provenance

Source

isl-org (Intel Intelligent Systems Lab)

Collection Method

Aggregated from various primary 3D research datasets and sensor outputs

Freshness

Last updated February 2026.

Geography

Global