KISAN SATHI – A Smart Crop Health Platform

KISAN SATHI is a project we are working on to help farmers identify crop problems at an early stage. The main idea came from a simple problem we noticed: sometimes a crop issue is noticed only after it has already affected a larger part of the field. We wanted to build something that can help the farmer understand the problem earlier and decide what action to take.

In our project, we are working with crop images to identify possible diseases or abnormal conditions. We have also worked with a chickpea disease dataset and trained an image classification model with different disease classes. The model gives a prediction along with a confidence score. Since real field images can be very different from training images, we don't consider every prediction as a confirmed result.

Along with the image, we are planning to use other information such as crop details, weather conditions, pest observations, sensor data and field location. The idea is to combine these details so that the system can give a better understanding of the condition of a particular field instead of depending only on one image.

We are also exploring how AWS can support the different parts of KISAN SATHI. Amazon S3 can be used for storing field images and other files, while DynamoDB can store information related to farmers, fields and crop observations. API Gateway and Lambda can be used for connecting the application with the backend. Since our idea also includes ESP32-based sensors, AWS IoT Core can be considered for handling sensor data. Cognito can be used for user authentication.

At the current stage, KISAN SATHI is a prototype and we are still working on improving different parts of it. We have not treated the AWS architecture as a fully deployed production system. There are still areas that need more testing, especially with different field conditions and real-world crop images.

Going forward, we want to improve the model with more diverse agricultural data and make the system more useful for actual field conditions. We also want to improve the integration of weather, sensor, location and satellite-based information. Our main goal is to make the system simple for farmers and help them notice crop problems earlier rather than only identifying them after major damage has happened.