The problem
Spray decisions depend on counting insects on sticky traps, which is tedious and subjective. Spraying on a calendar rather than on need raises costs and chemical load.
What to build
Create a tool that counts and classifies target pests from a photo of a trap and tells the farmer whether the threshold for spraying has been crossed.
Success looks like
Count error below 10% on a labelled trap dataset, and a threshold advisory customised per crop and growth stage.
Suggested technologies
- Object detection
- Few-shot learning
- Mobile camera
- Threshold rules
- Local language UI
An original, independent statement. It is not an official Smart India Hackathon problem statement, and Arventiq Labs is not affiliated with or endorsed by that programme. The scenario is illustrative and does not describe any specific organisation. Free to use with credit.