README.md

metadata

title: IndexRL
emoji: 🔥
colorFrom: indigo
colorTo: blue
sdk: gradio
sdk_version: 3.35.2
python_version: 3.8.9
app_file: app.py
pinned: false
license: mit

IndexRLApp

Gradio app to generate remote sensing indices for properties defined by image-mask pairs.

How to Install

IndexRL can be installed using the following two commands. However, it is recommended that you create a virtual environment prior to pip installation.

pip install -r requirements.txt
python app.py

Upload a Dataset

1. Set up the file structure of your dataset as follows.

.
└── dataset/
    ├── images/
    │   ├── img1.npy
    │   ├── img2.npy
    │   └── ...
    └── masks/
        ├── mask1.npy
        ├── mask2.npy
        └── ...

2. Create a zip file for the dataset

cd dataset
zip -r dataset.zip *

3. Go to the Datasets tab on IndexRL, upload the created zip file, give a name for the dataset, and press the Upload button.

Generate Remote Sensing Indices

1. Go to the Find Expressions tab, select the dataset from the dropdown and click the Find Expressions button. With time, more and more expressions will be generated.

2. Obtain the top 5 expressions from the Best Expressions box.

3. Visualize and debug the expression search tree from the panels on the right.

   • Iteration refers to each generated expression.
   • Tree Number refers to the index of the expression tree that resulted in the selection of each symbol of the expression.
   • Expression Tree displays the tree of symbols explored by the RL agent and the scores observed if each of those symbols were chosen. As you change the Tree Number, you can observe how the tree for the next choice of symbol.