Inspiration
Often fast response is needed to control the spread of diseases in a farm. Traditonal diagnosis methods relies heavily on experts which can be time-consuming, labor extensive. This platform solves the problem by providing quick detection of disease contracted plants during harvesting using AI.
What it does
This platform can classify upto 38 different classes of plant diseases by their leaf images. It also comes with a virtual assistant to solve any queries related to plant diseases which is powered by gemini.
How we built it
- A pretrained resnet50 model was fine-tuned on plant disease dataset.
- The virtual assistant is powered by google gemini.
- For frontend, Flask, a fast and lightweight python web framework is used and is hosted on google cloud run platform.
Challenges
It required meticulous parameter tuning and optimization to achieve optimal performance for pre-trained Resnet-50 model on plant disease dataset. Additionally, deployment on google vertex AI and integration with Flask frontend also posed technical challenges, requiring careful configuration and troubleshotting.
Accomplishments
I am proud that I was able to train as well as deploy the fine-tuned model which can classify upto 38 different classes of plant diseases on google vertex AI. Moreover, creating an integrated flask frontend, hosted on google cloud run which provides enhanced accessiblity and usability.
What we learned
Throughout this project, I gained valuable insights into various aspects of machine learning model development, deployment, and integration. Working with large-scale datasets, fine-tuning deep learning architectures, and deploying models on cloud platforms provided invaluable hands-on experience.
What's next for Plant Disease Classifier using AI
The next step for the Plant Disease Classifier using AI is to expand its capabilities beyond disease detection and question answering. The future version will integrate features like crop quality monitoring, efficient resource management, and more, utilizing IoT technologies and multimodal models. By incorporating various modalities such as images, sensor data, and descriptions, the aim is to enhance the robustness of the platform. Currently, the platform is limited to web, but I would like to extend it mobile platform by developing android/iOS application.
Log in or sign up for Devpost to join the conversation.