diffinytrace 2.4

No project description provided

DiffinyTrace

DiffinyTrace is a Python library for differentiable ray tracing and optical system optimization using PyTorch. It enables automatic differentiation through optical systems, making it possible to optimize lens designs, mirror configurations, and other optical components using gradient-based methods.

The source code is available at the GitHub repository.

Key Features

Flexible Transformations — apply general transformations such as rotations and translations to optical components, with full control over the parameters and their role in the transformation.

Seamless CAD Export — generate lenses and mirrors that can be exported to standard CAD file formats.

Freeform Surfaces — design complex optical elements with advanced B-spline representations for maximum flexibility.

• Differentiable Ray Tracing: Full automatic differentiation support through optical systems
• Constraint Optimization: Advanced optimization with PyTorch and SciPy integration
• Illumination Design: Algorithms for computing lens surfaces to achieve desired illumination profiles
• GPU Acceleration: CUDA support for high-performance computations

Installation

1. Create a new Enviroment via conda:

   conda create -n dit python==3.12
   

   activate enviroment via

   conda activate dit
   

   install pip

   conda install pip
   

2. Install PyTorch

   Check your cuda version with

   nvcc --version
   

   Diffinytrace only has been tested with 2.10.0+cu130. Make sure to install the appropriate version of PyTorch for your system. You can find the installation instructions on the PyTorch website. DiffinyTrace should work for both cpu and cuda versions.

3. Install DiffinyTrace Install all other dependencies and the library itself via:

   pip install diffinytrace
   

   or directly in the folder via

   pip install -r requirements.txt
   

Basic Usage Example

import diffinytrace as dit
import torch
NBK7 = dit.materials["NBK7"]

wave_len = 1.024
light_transform = dit.transforms.Offset(torch.tensor([0.0,0.0,0.0]))
source = dit.source.CollimatedMonochromatic(light_transform,8.0,wave_len)

plane_surface = dit.Plane()
surface2 = dit.Aspheric(-1/50.)
transf1 = dit.transforms.Distance(10.0,parent_transform=source)
lens1 = dit.Lens(transf1,5.,plane_surface,surface2,NBK7,13.0)
transf2 = dit.transforms.Distance(15.0,parent_transform=lens1)
detector = dit.Detector(transf2,plane_surface,8.0)
system = dit.SequentialOpticalSystem({"source":source, "lens":lens1, "detector":detector})

x,weights = source.sample(10)
O,D,wave_len,_,meta_data = system(x,["source","lens","detector"])
dit.plotting.system2D.plot(system,meta_data)

Documentation

For comprehensive documentation, tutorials, and API reference, visit the full documentation.

License

DiffinyTrace is licensed under the MIT License. See the repository for full license details.