https://github.com/opendronemap/skyremoval

Create sky masks to improve photogrammetric reconstruction
https://github.com/opendronemap/skyremoval

deep-learning image-processing onnx photogrammetry

Last synced: 6 months ago
JSON representation

Create sky masks to improve photogrammetric reconstruction

• Host: GitHub
• URL: https://github.com/opendronemap/skyremoval
• Owner: OpenDroneMap
• License: agpl-3.0
• Created: 2022-07-13T17:26:07.000Z (over 3 years ago)
• Default Branch: main
• Last Pushed: 2023-02-02T02:46:45.000Z (about 3 years ago)
• Last Synced: 2025-06-21T22:24:14.669Z (9 months ago)
• Topics: deep-learning, image-processing, onnx, photogrammetry
• Language: Python
• Homepage:
• Size: 41 KB
• Stars: 25
• Watchers: 7
• Forks: 3
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# SkyRemoval

The purpose of this tool is to create sky masks to improve photogrammetric reconstruction.

## Example

![img00471](https://user-images.githubusercontent.com/7868983/177582259-561c1e3a-529e-48a9-a122-2ca5b57a89d8.jpg)

![img00471_mask](https://user-images.githubusercontent.com/7868983/177581964-1f36ca2f-2d52-40e1-b4c2-6869b34ef2fc.png)

## Getting started

```

git clone https://github.com/OpenDroneMap/SkyRemoval.git

cd SkyRemoval

pip install -r requirements.txt

```

Usage:

```

python skyremoval.py source dest

```

It will automatically download the pre-trained model and run the processing. 

`source` can be a folder or a single image file. `dest` should be a folder.

## Parameters

```

usage: skyremoval.py [-h] [--model MODEL] [--ignore_cache] [--in_size_w IN_SIZE_W] [--in_size_h IN_SIZE_H] source dest

SkyRemoval

positional arguments:

  source                Source image path, can be a single image or a folder

  dest                  Destination folder path

optional arguments:

  -h, --help            show this help message and exit

  --model MODEL         Model path, can be a URL or a file path

  --ignore_cache        Ignore cache when downloading model

  --in_size_w IN_SIZE_W

                        Trained model input width

  --in_size_h IN_SIZE_H

                        Trained Model input height

```

## Using CUDA

If you want to speed up the processing, you can use CUDA. Check out the following link for more information:

https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html