An intelligent system for medicinal plant identification and symptom-based herbal treatment recommendation

Identify a medicinal plant from a photograph, describe how you feel in your own words, and explore the traditional herbal knowledge recorded for the match — with every score explained and every limitation stated.

How a result is produced

  1. 1

    Upload or describe

    A leaf photo, a symptom description in your own words, or both together.

  2. 2

    Real inference

    The image is classified and your text is scored against the stored herbal records.

  3. 3

    Explained output

    Confidence, similarity, matched terms and a Grad-CAM heatmap — plus the precautions that apply.

  4. 4

    Save or print

    Keep the analysis in your history or download a print-ready report with the safety notice.

Every figure on screen comes from the analysis service. If a result cannot be produced, the page says so instead of showing an example.

System modules

Four working modules, one integrated flow

Each module maps directly to a block in the approved high-level design, from user input through preprocessing and matching to the final system output.

System modules

Target architecture

The interface talks to a single API client, which calls the FastAPI service running the trained image classifier and the fitted symptom matcher over the project's curated datasets.

  1. Stage 1

    User interface

    React + TypeScript + Tailwind

  2. Stage 2

    API layer

    FastAPI service

  3. Stage 3

    ML & matching

    MobileNetV2 · TF-IDF + cosine similarity

  4. Stage 4

    Data & accounts

    Curated plant datasets · cloud database