Skip to content
This repository has been archived by the owner on Apr 2, 2019. It is now read-only.

Latest commit

 

History

History
43 lines (26 loc) · 1.6 KB

File metadata and controls

43 lines (26 loc) · 1.6 KB

Background-Separation-using-U-Net

Using a U-Net model to perform Image Segmentation on Carvana Image Masking Challenge Dataset

Setup

This project requires Python 3 to be installed.

If Python 3 is not default on the system, replace all pip with pip3 and python with python3

The python libraries needed are in the requirements.txt file. They can be installed using pip.

tensorflow-gpu is the default library. In case a compatible GPU is not available, change tensorflow-gpu to tensorflow in requirements.txt before installing using pip.

The following command can be used to install all the dependencies.

pip install requirements.txt

Once the dependencies are installed, the image data needs to be downloaded. The instructions to download the images are available in the respective folders in data folder.

Run

The file run.py can be run using the following command.

python run.py

The program takes an input for scale factor of the original image. Once finished, the best fit model and history of the run is exported to the current directory. The model can be loaded using Keras to further improve or for making predictions. The history can be used for visualization.

Report

The report of this project is available here.

License

This project is licensed under Apache License 2.0. The details can be found here.

Credits

All the contributions to this project is made by Aakaash Jois and Alp Aygar.