Computer Vision + NLP

Herbica AI

An AI assistant for medicinal plant identification that explains properties, safety notes, and usage guidance from an image-first workflow.

TensorFlowFastAPIReactOpenCVRAG

Problem Statement

Medicinal plant information is fragmented, hard to verify, and often inaccessible for everyday users.

Solution

A multimodal pipeline classifies plant images, retrieves structured herbal knowledge, and produces concise guidance with safety disclaimers.

Results

Prototype target: 95% class accuracy on curated species | Safety-first answer format | Fast image-to-insight flow

Future Improvements

Add regional language support | Add expert review mode | Deploy offline-friendly mobile capture

Architecture Workflow

01
Image upload
02
Vision classification
03
Knowledge retrieval
04
LLM response layer
05
Safety filter
06
User-facing report

Challenge Log

Handling visually similar species
Avoiding medical overclaims
Designing low-latency inference
Making outputs understandable
Metrics: 95% target accuracy | 200+ species dataset