vishalbdange / dryhunch-server

DryHunch

An AI/ML technology that forecasts droughts in specified places.

Team name: TechZombies

Team members

• Vriddhi Gupta - [email protected]
• Vedika Mamidwar - [email protected]
• Vishal Dange - [email protected]
• Shubham Khedekar - [email protected]

Description

• Drought is one such climatic disaster that occurs unpredictably and causes farmers to lose money owing to crop failure due to a lack of water.
• Farmers who have borrowed or invested in agricultural supplies such as seeds, fertilisers, and so on. Because of the money that they can't return, they commit suicide.
• Drying of water resources, crop failure, increase in food prices, poor health – and a decline in prices – of livestock are the most immediate impacts of drought perceived by farmers.
• As a result, we created an interactive and user-friendly mobile application that will help the farmers predicting the drought that may come in near future.
• It serves as a drought prediction system, forecasting forthcoming droughts in an area that poses a risk to farmers and reduce the risk of economic loss and social stress on them.
• It also advises farmers on drought related matters including drought declaration and suggesting ways to tackle the dry spell problem.
• After receiving the input area, the system will look it up on Google Maps, then acquire satellite photographs of it and continuously measure the green portion using Digital Image Processing and Convolution Neural Networks to get the predicted increase or decrease in the green part.
• The predicted percentage of green area in x days for an area will be combined with the other environmental indicators given in the first point to provide the final forecast of how many days the drought will last and the likelihood that it will.
• The Deep Learning Model is equipped with Weights & Biases to track, illustrate, and encourage collaboration during model training.

Links

• GitHub Repo link: Link to repository
• Flask Server: Link to flask server
• DL Model for predicting drought: Model code
• Application Working Video: YouTube Video
• Dataset Source: data
• Presentation Link: presentation

Technology stack

Tools and technologies that you learnt and used in the project.

1. Android Studio
2. Java
3. XML
4. Python Flask
5. PyTorch
6. Tensorflow
7. Visual Studio Code
8. Material UI

Usage

• After setup on Android studio is done, you can connect your mobile device to your laptop via USB and then run the application on your mobile. The app will be installed and opened automatically.
• When the app is finished, you will see a splash screen with beautiful design displaying the onboarding pages. After that, a series of onboarding screens will appear before you reach the home screen.
• No login is required because the application to everyone and no user data needs to be saved anywhere in anyway.
• On the home page, you will see various buttons and an edit text box. You can enter any location's name in the edit text box and click on "Get Prediction" which will take you to a page where you can view the satellite images of the specified location along with the probability of drought taking place in near future. Along with this, you can click on "Get Weather" button and get your current location's weather details.
• On the prediction page after entering location, when you click on "Get Suggestions" button, you will get to see advices about how to act depending upon the percentage of prediction.

Flowchart

Applications

• The system gives farmers advance notice of threats, which lowers the risk associated with drought impact. Additionally, it lessens the abrupt load of drought on farmers.
• Forecasting droughts is crucial for guaranteeing the security of both food and water since it has a significant impact on agricultural activity and water availability.
• Our application predicts drought in specific places in the next 10 to 30 days, but the meteorological department just forecasts rainfall in a specified area, and that forecast is not always reliable.
• It will help government agencies which focuses on drought and its impact on farmers.

Future scope

• Making the application an intuitive and multilingual Android app for farmers that allows them to input their current location in the form of speech or text.
• Increasing the accuracy of the existing Machine Learning Model as well as adding one more algorithm to predict in how many days a drought will possibly take place.

Screenshots

Splash Screen1	Splash Screen2	Instruction Page	HomePage

Prediction Page	Suggestions	Weather Details

• Deep Learning Model that will predict the drought condition