TL;DR: NeROIC generates high-quality 3D object models from online images, optimizing geometry and material properties.
NeROIC solves the problem of creating detailed 3D models from diverse online images, making it ideal for designers and developers needing high-quality object representations. It utilizes neural radiance fields to enhance geometry and refine camera parameters, providing accurate surface material properties and ambient lighting conditions for realistic rendering.
Download NeROIC: GitHub
Key Features of NeROIC
- Geometry Optimization: Enhances the scanned object’s geometry for more accurate 3D models.
- Camera Pose Refinement: Improves initial camera parameters for better image alignment.
- Surface Normal Extraction: Provides detailed normal maps to retain crucial object details.
- Material Decomposition: Solves for albedo, specularity, and roughness maps of the object.
- Relighting: Allows rendering of objects under novel lighting conditions.
How to Use NeROIC
- Collect images of the object from different angles.
- Use state-of-the-art methods to acquire camera poses and foreground masks.
- Optimize geometry and refine camera poses using NeROIC’s neural network.
- Extract surface normals and decompose material properties.
- Render the object with novel lighting environments.
Key Use Cases for NeROIC
- 3D Model Creation: Generate high-quality 3D models from online image collections.
- Novel View Synthesis: Create new perspectives of objects with accurate lighting.
- Relighting: Render objects under different lighting conditions for realistic effects.
- Material Analysis: Analyze and utilize material properties for detailed object representation.
