Inspiration
I belong to a place where most of the farming is done via word of mouth or based on ancestral beliefs, most farmers don't have the required logistics in order to do high level testing on their fields, however with the advent of wide-spread internet a lot of people have started using their phones to explore the world, I wished to use this in order to pay tribute to the various farmers who lost their crops due to bad planning or due to misinformation. I know that it isn't the greatest thing in this world, but I wanted to make my own contribution to some extent.
What it does
It is mostly a 2 page system, with one page working for Image classification for classifying images of plants with diseases the other one is a crop prediction based work.
How we built it
Most of the tasks were deployed using streamlit, we used a RESNet2 Architecture for the Image Classification task, the model was trained separataly on a notebook we then downloaded the pre trained model and then fine tuned it on the given dataset. The other task was predicting the best crop for a field given the conditions of the soil, the base model we used was a Decision Tree however a central problem was that if you are a farmer from a very backward place then you wont have any of that information so we decided to add a map feature where a person can select their location on the map and we would then use two api's one for finding the average weather and the other for finding the country. We decided to use an api for soil as well however soil grids is off the charts for now which led us to use an incomplete dataset for the task.
Challenges we ran into
Mostly the data finding tasks were very difficult as most of the datasets were incomplete also nearly all of the API's had a restriction so we also had to keep that in mind. Another issue was the integration of pytorch with streamlit as it caused a bit of the hassle
Accomplishments that we're proud of
We are really proud of the fact that we were able to identify a social problem and made a small contribution towards it.
What we learned
Mostly how to render stuff in streamlit, but I also peaked a lot into the Resnet architectures plus had to make a lot of trips to Kaggle as well just to understand the implementations as well.
What's next for Farm it Smart
We wish to create a more functional app along with having some other utilities but more importantly I wish to convert it to an android application because that is the way most of the people in my nation access the internet
Log in or sign up for Devpost to join the conversation.