svetadomoi / unet-rocks-segmentation

Simple U-Net model which is supposed to segmentate rocks in images

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Rocks segmentation with U-Net

This model is supposed to segmentate rocks in images made via flying drone. Since dice score is around 0.90 I can assure you that this model is pretty bad. My results you can see here down below.\

Original images:

Original images

True masks

True masks

Predicted masks

Predicted masks If you want to try it on your own follow this instructions:

Clone this repo

git clone https://github.com/svetadomoi/unet-rocks-segmentation.git

Install libraries

pip install -r requirements.txt

Download suitable GDAL wheel file from here (It's important not to change wheel file name), open cmd and navigate to the directory where you downloaded the file. Install it via

pip install path-to-wheel-file.whl

If you want to augmentate your images, do it via

python augment.py

After that you can train the model

python train.py

About

Simple U-Net model which is supposed to segmentate rocks in images


Languages

Language:Python 100.0%