This repository represents a web app with a multi-class classification ML model which creates a segmented image of rocks and plain land.

• This project is developed to solve the problem of detecting obstacles (eg. rocks) on lunar surface.

• Implementation is based on the U-Net architecture which creates a segmented image from raw image as an input.

• Artificial Lunar landscape dataset from Kaggle.

• Python 3.10.9
• PyTorch 1.12.1 (GPU)
• torchvision 0.13.1
• OpenCV 4.6.0
• Django 4.1.7
• cudatoolkit 11.6.0

Rest of the packages are listed in lunar packages list.txt file.

• Download the " final_model.pth " file from following Drive Link.
• Download the file and add its path in views.py file in load_checkpoint() function.

• Install the dependencies locally.

• To deploy this project open /lunarApp/views.py and run :

  python manage.py runserver

• It will launch the webapp, then follow below steps :

  1. Click on Choose File.
  2. Upload any file from Input samples eg PCAM1.png and click on segment.
  3. The results are displayed on new webpage.🎉🎊

• Paper 1
• Paper 2

If you have any feedback, please reach out to us at [email protected]