DeepLens is a differentiable optical lens simulator developed for:
Differentiable optical design
End-to-end optics-vision co-design
Photorealistic image simulation
DeepLens helps researchers build custom differentiable optical systems and computational imaging pipelines with minimal effort.
Note
This documentation is still under development. Some details may contain mistakes or be incomplete. We appreciate your patience and welcome any feedback or corrections.
Mission¶
Next-generation optical design software enhanced by differentiable optimization
Next-generation computational cameras integrating optical encoding with deep learning decoding
Key Features¶
Differentiable Optics: Leverages gradient backpropagation and differentiable optimization
Automated Lens Design: Enables automated lens design using curriculum learning and GPU acceleration
Hybrid Refractive-Diffractive Optics: Accurate simulation of hybrid lenses (DOEs, metasurfaces)
Accurate Image Simulation: Photorealistic, spatially-varying image simulations
Optics-Vision Co-Design: End-to-end differentiability from optics to vision algorithms
Contents¶
Getting Started
API Reference
- Lens API Reference
- GeoLens API Reference
- Class Hierarchy
- Main Class
- Initialization & Configuration
- Ray Sampling
- Ray Tracing
- Image Rendering
- PSF Calculation
- Optical Analysis (GeoLensEval)
- Geometric Calculations
- Lens Operations
- Optimization (GeoLensOptim)
- File I/O (GeoLensIO)
- Tolerance Analysis (GeoLensTolerance)
- Visualization (GeoLensVis)
- 3D Visualization (GeoLensVis3D)
- See Also
- References
- Optics API Reference
- Sensor API Reference
- Network API Reference
- Utilities API Reference