# sar_3D_multi_aspect

**Repository Path**: WshongCola/sar_3D_multi_aspect

## Basic Information

- **Project Name**: sar_3D_multi_aspect
- **Description**: The git repo for 3D reconstruction of paper Multi-baseline SAR 3D Reconstruction of Vehicle from Very Sparse Aspects: A Generative Adversarial Network based Approach
- **Primary Language**: Python
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 4
- **Forks**: 1
- **Created**: 2022-06-27
- **Last Updated**: 2025-03-17

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

The shared source code for benchmarking purpose of paper
# Multi-baseline SAR 3D Reconstruction of Vehicle from Very Sparse Aspects: A Generative Adversarial Network based Approach

---
Shihong Wang, Jiayi Guo, Yueting Zhang, Yirong Wu, in ISPRS-journal-of-photogrammetry-and-remote-sensing

![Graphical Abstract](./Graphical_Abs.png "Graphical Abstract")

---

## Data
Simulation Data (Civilian Vehicle Demo):

https://www.sdms.afrl.af.mil/index.php?collection=cv_dome


Measured Data (GOTCHA):

https://www.sdms.afrl.af.mil/index.php?collection=gotcha

---
## Requirements

Python == 3.7

numpy >= 1.12.1

tensorflow >= 2.0.0

---
## Run
Please firstly run the files in the CSImaging dir to make spacial datasets from the Simulation or Measured Datasets.

<code data-enlighter-language="raw" class="EnlighterJSRAW">Python CSImaging/01_Extract.py</code> to extract an isolated target from the entire scene.

<code data-enlighter-language="raw" class="EnlighterJSRAW">Python CSImaing/02_GridInterp.py</code> to make interpolation in the frequency domain for the convenience of FFT.

<code data-enlighter-language="raw" class="EnlighterJSRAW">Python 03_CS.py</code> to run the imaging algorithm described in the paper.

<code data-enlighter-language="raw" class="EnlighterJSRAW">Python 04_Noncoherent_Max.py</code> to accumulate the imaging results of several aspects as the input or ground truth of network training and validation.

Then, Run

<code data-enlighter-language="raw" class="EnlighterJSRAW">Python train-model.py</code>

to train the network and save the weights.

Finally, run

<code data-enlighter-language="raw" class="EnlighterJSRAW">Python view-pred.py</code>


to see the predictions of network from few aspects.

Generally, based on the training configuration and steps explained in our paper, the network will be trained for benchmarking purpose.

---
## Citation

If you use this paper, network and imaging code, 
please search the paper on the website of ISPRS and cite this paper.