AgroAI is an intelligent agricultural assistant that detects plant diseases from a single leaf photo. Built using Vision Transformers trained on the PlantVillage dataset via HuggingFace. Designed for farmers in developing countries who cannot afford lab testing. Upload a clear photo of a leaf and click the button to have the system examine it. The app uses AI to check the leaf for signs of disease or stress and returns an easy‑to‑read health report for you...
No reviews yetBe the first to leave a review for AgroAI-Diagnostic-Center
Maker
📌
Hey Product Hunt community! I’m Chikamso, the maker of AgroAI, and I’m incredibly excited to share this project with you all today.
What inspired me & the problem I’m solving:
While working with advanced AI vision models, I noticed a significant gap: they often struggle with regional contexts. When trying to diagnose crop diseases for West African flora, standard models would frequently hallucinate or misidentify local plants. I wanted to build a localized, mobile-first diagnostic tool that farmers and agricultural experts could actually rely on in the field without needing a heavy app download.
How the process evolved:
Building this was a massive learning curve in managing AI bias and prompt engineering. Initially, I hardcoded the system to look for specific regional plants like the African Pear (Ube). However, this created a bias loop where the AI started identifying every leaf as an African Pear!
I had to completely re-architect the backend to use a “Neutral Expert” prompt—forcing the Llama-4-Scout model (via Groq) to rely strictly on visual botanical evidence (venation, margins) before applying its regional knowledge.
As the project evolved, I realized a one-off scan wasn’t enough. I upgraded it into a full professional suite by integrating a local SQLite database for persistent scan history and building a WhatsApp bridge so users can instantly forward Markdown-formatted diagnostic reports to experts.
Tech Stack: Flask, Python, Groq API (Llama 4 Vision), Tailwind CSS, SQLite, hosted on Hugging Face Spaces Docker containers.
I’d love for you to test it out! Upload a photo of a plant leaf and see how it performs. I’ll be hanging out in the comments all day to answer questions and hear your feedback.