CropGuard
Point your phone at a sick plant and know what’s wrong in seconds.
What I’m building
CropGuard is an open-source app that identifies crop diseases from a single photo and tells the farmer, in plain Spanish, exactly what to do next — organic first, chemical last. It’s trained on the diseases that actually hit smallholder coffee, cassava, and maize, not the ones in a European textbook.
Why I’m building it
My family farms coffee. One bad season of leaf rust wipes out a year. Agronomists exist, but not on a Tuesday afternoon three hours up a mountain road. A phone is already in every pocket up there. The knowledge should be too.
What the tokens unlock
Credits fund the GPU time to fine-tune the vision model on our field-collected dataset, plus the model calls that turn a diagnosis into step-by-step, region-specific advice a farmer can act on today.
- 1Fine-tune the disease-detection vision model on field photos1.2M
- 2Build the plain-language treatment advisor800K
- 3Field-test with 200 farmers across 3 regions1M
The builder
Medellín, Colombia 🇨🇴
Agronomist + self-taught ML tinkerer. My family farms coffee; I want every smallholder to have a plant doctor in their pocket.
Updates
Dataset crossed 12,000 labeled photos
Farmers themselves are labeling now. Accuracy on leaf rust is already at 91% on held-out images.
Supporters
18 people have pledged so far.
- 500KOpenAIKen WatanabeTokyo, Japan· 2d ago
“Rooting for you. This should exist.”
- 250KComputeMaría TorresBarcelona, Spain· 4d ago
- 1MClaudeAnonymous supporter· 6d ago
“Take my credits and go build it.”
- 500KGeminiAnonymous supporter· 8d ago
- 100KAI CreditsAnonymous supporter· 10d ago
Discussion
Ask the builder a question or leave encouragement.
- No comments yet. Be the first to say something.