Automated 3D Object Classification with Multi-View Rendering
Previously, we introduced a segmentation system that required manually assigning names to objects using the “semantic_name” attribute. Now, with an AI-powered feature, that process is automatic. Using multi-view rendering and OpenCLIP, the system classifies 3D objects and assigns meaningful names without manual input.
Key Features:
- Multi-View Rendering: Renders objects from different angles for better classification.
- AI-Powered Automation: Automatically assigns semantic names to objects based on the AI model.
- Customizable: You can integrate your own classifier to improve accuracy.
Why It’s Useful: This tool is ideal for unclassified 3D scenes used as synthetic data. It helps prepare objects for segmentation quickly and efficiently, reducing manual effort.
Explore the feature in my repository: GitHub Repository.
The relevant test is in object_classifier_test.py.
Learn more about synthetic data in my previous post here.
Scene credits :
Dekogon Studios – Props