Computer vision · Android

KU Krishi

An offline-first Android app that diagnoses crop diseases such as late blight from a photo of a leaf, on the phone and with no signal needed, then explains the result in Nepali.

Year
2024 · Kathmandu University
Recognition
Nepal ICT Award 2024, semi-finalist
Platform
Android · published under bala.ai

01 Problem

Farmers in remote areas lose crops to diseases like late blight before an agronomist can ever see them.

Many of those farms have weak or no mobile signal, so a diagnosis tool that depends on the internet fails exactly where it is needed. KU Krishi was built to work with no connection at all.

02 How it works

  1. Leaf photo
  2. On-device YOLOv8
  3. Disease diagnosis
  4. Advice in Nepali
Fig. 01 — Inference runs on the phone, so a diagnosis never waits for the network.
  1. Leaf photo

    The farmer photographs an affected leaf with the phone camera.

  2. On-device model

    Quantized YOLOv8 and MobileNet models run on the phone with TensorFlow Lite, so nothing is sent over the network.

  3. Disease diagnosis

    The app identifies the disease, such as late blight, from the photo.

  4. Advice in Nepali

    The result and what to do next are explained in Nepali, the farmer’s own language.

03 Result

  • 2024

    Semi-finalist, Nepal ICT Award · built at Kathmandu University

  • Live

    A published Android app farmers can use with no signal · bala.ai

04 Stack

Mobile

  • Flutter
  • Firebase

Machine learning

  • YOLOv8
  • MobileNet
  • TensorFlow Lite
  • OpenCV

Backend

  • FastAPI

05 Next project

Hydroponics AI

Reinforcement learning · IoT · STEAM research grant