Inspiration

The agricultural sector is considered an essential industry in the current COVID19 pandemic. However, the pandemic has shown several outstanding issues that have not been effectively resolved, such as:

  • Most agricultural products come from small and medium-sized enterprises that do not qualify for export. Agricultural by-products have not been used effectively.
  • Due to social distancing, it is difficult for farmers to get direct advice from experts.
  • Finally, knowledge and experience are also a major barrier to entering the industry, especially for young businesses.

What it does

The mobile app helps monitor and protect crop health and supports the transition to the circular economic model with AI technology for small and medium businesses operating in the agricultural sector.

How we built it

Setup Project: Goal: Have an application can show result from a captured image Performance constraints: Have a high speed and good accuracy

Data Pipeline: Data availability in Kaggle Storage: Google Driver, Azure Storage Exploratory Data Analysis: Checking and get insight from data (Python, Matplotlib, Pandas,...)

Modeling: Model selection: First choice Densenet, Resnet50, MobileNet Training and Debug (Python, Tensorflow, Keras)

Serving: Building back end by Flask Using Azure storage to save Model Using Azure Ubuntu Virtual Machines for Server (building API for Deploy app)

Deploy app: Using Java with support of android studio to build file apk for Hoori App.

Challenges we ran into

  1. AI Recognition System:

Besides the searching function by text, the AI ​​system recognizes and diagnoses the disease, then provides information, Solutions, and medicine. Currently, our model is over 90% accurate and will be higher if we have more data. Especially, Hoori focuses on organic solutions to treat diseases to limit chemical pesticides and make use of what's already there. If these ways are not efficient, we introduce the pesticide with low chemical content and friendly with the environment. Users can order right on the app because Hoori is linked with e-commerce sites.

  1. The plant disease monitoring system:

Users can create track records for a certain crop with a specific code. Then each time using the AI recognition function, the user just needs to enter that code. The system will evaluate the crop's status based on the photos taken by the user in the past and at the most recent.

  1. The information proposal system:

Be customized by Machine Learning. From the data we already collect from a user, Hoori will propose relevant information about the circular economy that users can apply based on what users have. For example, this person is growing an apple tree, and we can suggest ways to utilize the wastes from this tree planting. This approach can help users save costs, generate more profits, and increasing users' knowledge and awareness of the circular economy.

  1. The integration function:

In the future, we want to integrate our AI model into farm cameras or cameras from drones to detect diseased crops quickly. Since then, there will be timely and effective support measures. And in the future, we plan to build a website version to approach more customers.

Accomplishments that we're proud of

We had to build a machine learning flow by myself This is the first time we deploy the model to an application, and we have worked hard, so in the end, the app was done, and get as much as our target. We have more time to get Experiences with Azure, and we think this is a good foundation for AI's future.

What we learned

Using Azure foundation for deploy an AI application Survey data and solve the problem

What's next for Hoori

  • Looking for a team to complete
  • Build a startup
  • Making an MVP

Built With

Share this project:

Updates