Inspiration
We recently went to a village, and to our surprise, we didn't know the names of any crops. We could not identify them. So we came to a realization that the youth is forgetting our farming culture. We decided to do something about it.
What it does
The user simply points the camera to the crop and the app identifies it and provides the user with a description and details. Farmers can also point the app to crops and identify whether the crop is disease infected. That'll help them to find suitable measures to prevent further damage.
How we built it
We began with dataset collection. Then we trained a model to identify different crops & the infected crops. We then converted the model to tensor flow lite to use it in our Android app. and Finally we built the app with this offline tflite model.
Challenges we ran into
- Finding the dataset. we scraped a lot of websites for images and datasets. This was the hardest part.
- Training the model. Finding out the optimal iterations and parameters took some time.
- Kotlin. None of us knew kotlin. So it took us some time to get started with the app.
Accomplishments that we're proud of
Successfully built a working demo for the flower database.
What we learned
- How to train a custom tensor flow lite model.
- Working with kotlin.
- Scraping datasets and images.
What's next for TPA : The Plant App
- Crop health Prediction based on current state of crops.
- Integrated Google Maps to directly upload field pictures and serve as a local guide.
Built With
- firebase
- kotlin
- tensorflow-lite
Log in or sign up for Devpost to join the conversation.