Ommatdium

https://www.youtube.com/watch?v=a7IYtKkpXR4

https://github.com/raghavnautiyal/Farm-Free

Inspiration 😥

Farmers rely heavily on crop yields as their main source of income. However, many pests can get in the way of optimal production yields as they can eat and harm the field crops. Most of the farmers are not able to recognize the bug and whether it's harmful to the crop or not. And with that too they don't know how to tackle the bug. The most common method to get rid of them is pesticides, but different pesticides attack different bugs.Despite that, the large quantities of pesticides used to spray on acres of farmland can be expensive, and many times, farmers misuse pesticides on the wrong bugs. The overuse of pesticides has many consequences which can cost farmers money, time, and resources:

• First, the EPA estimates up to 70 million pounds of pesticides are lost to drift each year, a common issue in which extra pesticide chemicals are carried by the wind, hurting the ecosystems, the farmer's wallet, and human health.
• Second, an overabundance of pesticides on the wrong species can lead to pesticide resistance. As a result, pesticide costs should be expected to increase as new variations of the pesticide are more expensive.
• Third, spraying pesticides on beneficial pests can negatively impact the production rate of the farmer's crops, which is a waste of money.

These effects worsen each year, causing rural/small farmers to lose thousands of dollars, as on average, they spend around $21,000 on pesticides alone as of 2019. This number will only increase in the coming years.

So there's an immense misunderstanding to use the right pesticide to get rid of the right bug.

We brought a long-needed solution for this and we called it Ommatdium.

What it does 😎

Ommatdium is a machine learning-based full-stack application that will help farmers identify harmful bugs for their crops and will also suggest to them the proper way to get rid of the bugs/pest. Ommatdium can also predict future bug swarms, influxes or migrations. This app would be the best buddy to a farmer or even a normal person who likes to plant small plants at his home.

The two main features of Ommatdium

1. BugScan - Users are allowed to upload an image of a bug on the plant. Then Ommatdium identifies the bug using machine learning. After analyzing and matching the bug-pest database it suggests whether it's harmful to your crop or not. If its harmful Ommatdium also suggests which pesticides to use, to get rid of them without harming the rest of the plant/soil.
2. Heatmap - Ommatdium also has a heatmap feature where users can see predicted future bug invasions according to their location. The app also maps a harmful bug report on the heatmap and crowdsources the information so that other users in a 2km radius are notified about it. Because if they can have it you can have it too. This feature increases the mutual benefit.

How we built it 😙

Building this app was more complex than we initially thought. We discuss how we made it in a stepwise manner :-

1. Frontend - We coded the frontend in HTML, CSS, javascript and Semantic UI. We used leaflet.js for adding the map feature on our platform.
2. Backend - We have flask as our backend server which is connected with the frontend. We use firebase as our database and the ML model is also integrated within it. The app gets the user's location firstly then When the user uploads image, it gets stored along with the location into firebase and at the same time is run through the model, the model returns the bug_name parameter which is then checked within the bug-pest database and then returned on to the frontend with bug_name, harm_or_not, description, pest_name as the returned values. If the bug is proven to be harmful we extract the location from the firebase and display it on the map.
3. Machine learning model - We found a dataset of 102 most common pests and trained our model in TensorFlow. After training it for over 4 hours we got an accuracy of 94%. Then we integrated the model into the flask server which communicates with the backend. It intakes bug_image and returns bug_name from and to the backend
4. Heatmap/future prediction - Using the data the farmer enters and combining with NASA Terra Sattelite vegetation data we will create a heatmap to predict where pests influxes will most likely be travelling next based on what plant they are inclined to attack. Our app will warn the person according to their location
5. Currently, the app is hosted locally we plan to soon host it with Heroku and are halfway done with it but couldn't complete due to time constraints

Ommatdium benefits scientists, farmers, and nations alike. Allowing that 40% of crops to be available for people facing food insecurity. Providing bug Migration Data for Scientists. Creating Safer food for the entire world. We hope to increase biodiversity, and better prepare farmers for incoming pests.

We enjoyed building this app. Each and every line of code is written by us

The main point where we used NASA data was in our map feature. We used the NDVI layer from NASA's MODIS satellite and overlayed it on our map. The purpose is for users to see the different concentrations of vegetation and explore different correlations between the vegetation and where the bugs are most likely to go next. We plan to write code that will update our data on the map every day and make it more realtime.

Over this weekend our team learned a lot. Through Ups and Downs, our team persevered. We pivoted but we never gave up. We learned new skills we never had. Going through this experience created bonds with our teammates. We had lots of laughs and also lots of game faces. Looking at the world we saw so many people facing food insecurity which made us think about how we can help the farmers produce more food and the number one problem is pests. Our team was inspired to find a solution to give back. We wanted to build a solution that would enable farmers and scientists to be prepared for what comes next. We had a challenge figuring out what frameworks to use on our site and what would mesh well with our ML aspect which we resolved by consulting mentors. We'd like to thank all the Mentors and Judges who have helped us this weekend.

Dataset for training model, https://www.kaggle.com/kmldas/starter-kernel-insect-identification/notebook

Bug Classes: https://docs.google.com/spreadsheets/d/1Xc4yWOaHbKskEMvhLQU4qVLCVCAzyzxQtePf7e7CW1U/edit?usp=sharing

https://neo.sci.gsfc.nasa.gov/view.php?datasetId=MOD_NDVI_M

Tools: Tensorflow, Pandas, Python, Visual Studio Code, Heroku App, NDVI layer from MODIS satellite (NASA), HTML,CSS,JS, Semantic UI, firebase and leaflet.js

#software #bugs #farmers #ML #biodiversity

This project has been submitted for consideration during the Judging process.