3D Visual AI Guide#
Complete 3D Data Workflow with Point Clouds, Meshes, and 3D Annotations
Level: Intermediate | Estimated Time: 20-30 minutes | Tags: 3D Data, Point Clouds, LiDAR, 3D Meshes, 3D Annotations, Spatial Data
This step-by-step guide will walk you through a complete 3D data workflow using FiftyOne. You’ll learn how to:
Load and visualize raw 3D data including point clouds and meshes
Work with different 3D file formats (GLTF, OBJ, PLY, STL, FBX)
Add 3D annotations like bounding boxes and polylines
Navigate and interact with 3D scenes in FiftyOne’s viewer
Organize 3D datasets for spatial analysis and model training
Guide Overview#
This guide is broken down into the following sequential steps:
Getting Started with 3D Datasets - Learn how to load and visualize raw 3D data, including point clouds, inside FiftyOne, exploring basic navigation in the 3D viewer and organizing datasets for spatial tasks
Loading 3D Annotations - Take your 3D workflows further by adding annotations like bounding boxes and labels to your point clouds, learning how to bring in annotations and overlay them seamlessly for inspection and validation
Prerequisites#
Who Is This Guide For
This guide is for developers and computer vision engineers looking to work with 3D datasets in FiftyOne. Whether you’re dealing with LiDAR point clouds, voxel grids, or 3D bounding boxes, this series introduces the tools and workflows you’ll need to visualize and curate your 3D data effectively.
Required Knowledge
You should be comfortable with the FiftyOne dataset structure and basic computer vision concepts. Prior experience with point cloud data or 3D annotations is helpful, but not required.
Packages Used
The notebooks will automatically install the required packages when you run them. The main packages we’ll be using include:
fiftyone - Core FiftyOne library for dataset management and 3D visualization
open3d - 3D data processing and point cloud operations
numpy - Numerical operations and array manipulation
matplotlib - Visualization and plotting
Each notebook contains the necessary pip install commands at the beginning, so you can run them independently without any prior setup.
System Requirements
Operating System: Linux (Ubuntu 24.04), macOS
Python: 3.9, 3.11
Memory: 8GB RAM recommended for 3D operations
Storage: 5GB free space for 3D datasets
Graphics: GPU recommended for large point cloud visualization
Notebook Environment: Jupyter, Google Colab, VS Code notebooks (all validated)
3D Data Formats#
Point Clouds
LiDAR Data - Raw sensor data from autonomous vehicles and robotics
PLY Format - Standard point cloud format with color and normal information
PCD Format - Point Cloud Data format for efficient storage
XYZ Format - Simple coordinate-based point cloud representation
3D Meshes
GLTF/GLB - Modern 3D format with materials and animations
OBJ - Wavefront Object format for 3D geometry
PLY - Stanford Triangle format for 3D meshes
STL - Stereolithography format for 3D printing
FBX - Autodesk format for 3D content exchange
3D Annotations
3D Bounding Boxes - Object detection with location, dimensions, and rotation
3D Polylines - Path and lane annotations in 3D space
3D Segmentation - Point-level annotations for semantic understanding
3D Data Workflow#
This tutorial demonstrates a complete 3D data workflow that combines:
Data Loading - Loading different 3D formats (point clouds, meshes) with proper coordinate systems and transformations
Scene Management - Creating and organizing 3D scenes using FiftyOne’s Scene class and FO3D file format
3D Visualization - Navigating and interacting with 3D data in FiftyOne’s interactive 3D viewer
Annotation Integration - Adding 3D bounding boxes, polylines, and other annotations for comprehensive analysis
This integrated approach gives you the tools to not just load 3D data, but to understand spatial relationships, manage complex 3D scenes, and prepare datasets for 3D computer vision applications.
Ready to Begin?#
Click Next to start with the first step: Getting Started with 3D Datasets in FiftyOne.