KNUST_MedTechCrushers

We present an easy-to-use mobile app for farmers to detect diseases in tomato and cocoa plants. The app also suggests the appropriate remedy so that the farmer can take timely action.

Comment

Inspiration

Coming from a country where agriculture is a primary occupation. this also ahs impact in my country.

What it does

Our app captures an image of either a tomato leaf or cocoa plant, sends the image as a HTTP request to azure cloud and provides inference results

How we built it

We built our model in a pipelined fashion. The stages of detection are step1: User takes a photo of the diseased plant with their phone step2: The photo is uploaded to our machine learning model hosted on the azure cloud. Step3: The model detects the plant disease and sends a response informing the farmer of the type of disease and possible remedy.

We have trained our model using Keras, using a resnet50 backbone.

Challenges we ran into

Our original solution was to allow the user to send an MMS to a an MMS/SMS gateway server that could then upload the sent MMS to the Azure cloud. The predicted disease and the remedy would then be sent back to the farmer's mobile as a simple SMS response. We opted for this solution because most farmers in Ghana might not be able to afford a high-end smartphone with internet connectivity or a good enough RAM to run machine learning models.

We tried to use Twilio to host an MMS server but were unsuccessful as MMS services are available only if the server location is set to the US. It was not a practice to use an MMS server hosted in the US, because then the farmers would have to send MMSs overseas, making the service very expensive and thus unfeasible.
However, if a dedicated MMS server is hosted in Ghana then our solution would work.

Accomplishments that we're proud of

What we learned

We learnt how to build and deploy machine learning models in the Azure cloud. Additionally, we also learnt how to build a mobile application using swift and use it to send data to the Azure cloud

What's next for KNUST_TUM_medtechcrushers

The next focus of KNUST_TUM._medtechcrushers would be to

  1. create an app that runs the disease classification model ondevice.
  2. potential collaboration with MMS/SMS gateway provider to allow local us to host an MMS server in Ghana so that our service can work on very low-end phones.

Built With

Share this project:

Updates