med-el-ghazoui/plantcare-ai

How med-el-ghazoui/plantcare-ai works

AgriScan AI is a preliminary plant-disease screening product: an image reaches the API, is validated and classified, and the caller receives a prediction or uncertainty result. Authenticated users may have analysis metadata saved; a separate endpoint returns an annotated image. The repository also contains model training and evaluation, including external field-validation tooling. The README documents the intended mobile flow, but the sampled mobile files are platform shells; field evaluation is supported by the source index and excerpts, not a full inspection of its subsystem.

Main entry points: main.py FastAPI routes · auth.py Account auth · exports.py Annotated export · field.py Field evaluation

Components 23

API and inference

FastAPI routes backend/app/main.py
Image validation backend/app/image_validation.py
Model inference backend/app/inference.py
Confidence decision backend/app/decision.py
Annotated export backend/app/exports.py
Image watermark backend/app/watermark.py

Accounts and history

Account auth backend/app/auth.py
Credential security backend/app/security.py
Login throttling backend/app/throttle.py
Analysis history backend/app/history.py
Database access backend/app/db.py
User and analysis records backend/app/models.py

Model development

Dataset pipeline ml/agriscan_ml/data.py
CNN architectures ml/agriscan_ml/models.py
Model training ml/scripts/train_mobilenet.py
Classification metrics ml/agriscan_ml/metrics.py
Selective prediction ml/agriscan_ml/selective.py
Model evaluation ml/scripts/evaluate.py

Field validation

Field manifest checks ml/scripts/validate_field_manifest.py
Field evaluation ml/agriscan_ml/field.py
External evaluation ml/scripts/evaluate_external.py

Other components

Image user
Mobile shell mobile/android/app/src/main/kotlin/ma/agriscan/agriscan_mobile/MainActivity.kt

Connections 28

Image user → FastAPI routes: uploads image README.md
Image user → Mobile shell: uses client README.md
Mobile shell → FastAPI routes: sends image README.md
FastAPI routes → Image validation: decodes image backend/app/main.py
FastAPI routes → Model inference: runs prediction backend/app/main.py
FastAPI routes → Confidence decision: applies threshold backend/app/main.py
FastAPI routes → Account auth: checks credentials backend/app/main.py
FastAPI routes → Database access: stores analysis backend/app/main.py
FastAPI routes → Annotated export: registers routes backend/app/main.py
Annotated export → Image validation: decodes image backend/app/exports.py
Annotated export → Model inference: runs prediction backend/app/exports.py
Annotated export → Confidence decision: applies threshold backend/app/exports.py
Annotated export → Image watermark: composes output backend/app/exports.py
Account auth → Credential security: verifies credentials backend/app/auth.py
Account auth → Login throttling: checks login limits backend/app/auth.py
Account auth → Database access: loads user backend/app/auth.py
Account auth → User and analysis records: creates user backend/app/auth.py
Analysis history → User and analysis records: queries analyses backend/app/history.py
Analysis history → Account auth: requires identity backend/app/history.py
Model training → Dataset pipeline: loads batches ml/scripts/train_mobilenet.py
Model training → CNN architectures: builds classifier ml/scripts/train_mobilenet.py
Model evaluation → Dataset pipeline: loads split ml/scripts/evaluate.py
Model evaluation → Classification metrics: reports scores ml/scripts/evaluate.py
External evaluation → Field manifest checks: validates manifest ml/scripts/evaluate_external.py
External evaluation → Field evaluation: evaluates predictions ml/scripts/evaluate_external.py
Field manifest checks → Field evaluation: checks protocol ml/scripts/validate_field_manifest.py
Field evaluation → Classification metrics: computes metrics ml/agriscan_ml/field.py
Field evaluation → Selective prediction: uses threshold metrics ml/agriscan_ml/field.py

Drawn by GitDiagram on November 10, 2026 from the repository's file tree, README and source files. Read as Markdown ·