ClimateAI-CAELI

MVP

Problem Statement:

Solar power will play a major role towards sustainable energy transition. Reliable operation of solar power plants is critical both in terms of power system safety and economics. The use-case we aim to solve here is to improve the operational performance of solar modules enabling focused maintenance and improved efficiency. This is a multi-classification problem using infrared images where our aim is to classify whether a given image belongs to the healthy category or one of the unhealthy ones.

Dataset:

https://github.com/RaptorMaps/InfraredSolarModules

Folder structure:

1. modelTraining: It contains scripts used for model training and saving.

2. prediction: It contains code for predicting the class given a new image.

3. static: style sheet, images and model metadeta for web application.

4. templates: HTML for web application.

5. image: contains any image used either in readme

6. project-planning: Project Planning and Version Control workflow management.

Hardware detail:

This project is developed on Windows 10 operating system without any GPU.

Setup:

Step 1: Create virtual environment

conda create -n hiveProject python=3.6
conda activate hiveProject

Step 2: Install required library to the virtual environment from requirements.txt file.

Step 3: Train model using CNN notebook in modelTraining folder.

Step 4: Save the model and other encoding parameters in static/model-metadeta folder for the web application to pick it.

Step 5: Run app.py to check the application functionality in development.

python app.py

Step 6: Create a new app in heroku and push the code to heroku master