CGH-SFO-solver

KO | EN

gitlite — search

#hacktoberfest #ai #golang #android #machine-learning #deep-learning #linux #macos #windows #kubernetes #ios #mcp #openai #open-source

(★ 16)

Propagation-adaptive 4K computer-generated holography using physics-constrained spatial and Fourier neural operators

CGH-SFO-solver 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# CGH-SFO-solver (Updated 03/19/2025)
Propagation-adaptive 4K computer-generated holography using physics-constrained spatial and Fourier neural operators  
A copy that contains the reported data in the manuscript can be accesssed from https://drive.google.com/drive/folders/17h8pox1Wh5M2rPspZ6HLve3HG38B9BC0?usp=sharing  
Already submitted to Nature Communications; Submission date: 16/12/2024  
Current state: major revision. Hope to meet the public soon!  
(Model weights are curretly not open to public since the paper is under review; Results and codes can be accessed)

## Structure and usage   
── `model/`: trained models for r(638nm), g(520nm), b(450nm) channels (FourierNet_flex_100.pth)  
Model weights https://drive.google.com/drive/folders/1p6oGe6SAp2JGSmdL7CtNRSu_-bFknfKS?usp=sharing

&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;── `FourierNet.py`: contains modules of SFO-solver.  
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;Input: targeted intensity & targeted distance (85mm-115mm); Output: phase-only hologram

── `testdata/`: testdata that readers can try out on, criterion calculation(PSNR&SSIM) can also be found in the link  
Data https://drive.google.com/drive/folders/1QnJ3ify3CK2_eTX7T35ZY2TcpX4eEVRM?usp=sharing

── `Dual_plane/`: Data for dual-plane 3D holographic projection that is shown in the main manuscript Section 2.5 and Supplementary Note 5.
https://drive.google.com/drive/folders/1dqgYg9yUl70pzBY6hTmSWPMTrXKDplBr?usp=sharing

--------------------------- **Evaluating SFO-solver performance with trained model weights** -----------------------------  
── `eval_test.py`: Run SFO-solver and simulate reconstructions with provided weights on single-plane test data.  
```python
# Traverse all test data across three wavelengths and 85-115mm distances with 1mm interval
python eval_test.py
```

── `eval_dual.py`: Run SFO-solver and simulate 3D reconstructions with provided weights on dual-plane test data.  
```python
# Place the two objects at 85mm and 115mm respectively
# name indicates the object to choose
python eval_dual.py --channel=0 --name=symbol --distance=[85,115] # red:0, green:1, blue:2
```

── `compute_criterion_testdata.py`: PSNR and SSIM calculation for test data.  
```python
python compute_criterion_testdata.py
```

---------------------------------------------------- **Training SFO-solver** ------------------------------------------------------  
── `flex_train.py`: Main code to train SFO-solver based on self-supervised learning.  
```python
# Train on channel 0 (red)
# Sample distance from 85mm to 115mm
# Use the merge loss mentioned in the manuscript
python flex_train.py --channel=0 --distance=[85,115] --loss=Merge2
```

**Utility functions**  
── `loss.py`: Loss functions for training

── `prop.py`: Simulated propagation functions

── `dataset.py`: Dataset preparation for training

── `distance_generation.py`: Randomly sample distances to compose dataset

── Algorithm_comparison: Running code for DPAC, SGD, and GS algorithms   
Modified from https://github.com/computational-imaging/neural-holography

CGH-SFO-solver

CGH-SFO-solver 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

CGH-SFO-solver

CGH-SFO-solver 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README