Yahia Fares University • PFE
AgroVision9
API Online
Active Inference Model
Potato Disease
YOLO detector for image, video, webcam, and drone workflows.
Model Expansion

Multi-model workbench

Select a YOLO detector or ONNX classifier, check whether the artifact is loaded, and keep each model family separated from the primary workflow.

Loading model catalog...
Leaf Triage

Leaf diagnosis with side-by-side proof

Upload one plant leaf image and compare the raw input with the annotated result, confidence bars, and a download-ready report image.

Leaf
Drop a plant leaf image
JPG, PNG, WEBP, or BMP. Best for plant disease diagnosis and field demos.
Uploading...
Original Image
Original leaf
YOLO Annotated Result
Annotated result
Download Result
Sequence Review

Video analysis for batch inspection

Upload field footage to estimate the dominant disease class across frames, review class distribution, and export the annotated MP4.

Drop a field inspection video
MP4, MOV, or AVI. Maximum file size is 100MB.
Uploading...
Frame distribution timeline
Annotated Video Output
Download Video
Live Monitoring

Realtime webcam inference

Stream local frames into the model and watch disease class, confidence, latency, and overlay boxes update live.

Camera feed appears here when live inference starts.
Aerial Operations

Drone mission workspace

Queue demo-safe commands, monitor telemetry from the bridge app, and overlay YOLO detection on the incoming RTMP stream.

Connect your drone via RTMP
Point any drone/phone live-stream app at this server, then watch the YOLO overlay below.
rtmp://159.89.28.191:1935/live/drone
  1. Power on the drone and connect your phone or controller to it.
  2. In your live-stream app (DJI Fly, Litchi, or any “Custom RTMP” option) choose RTMP / custom streaming.
  3. Paste the server URL above as the stream address and start the broadcast.
  4. AgroVision9 reads that stream internally as rtmp://mediamtx:1935/live/drone (already filled in below) — just press Start Stream.
  5. The annotated feed appears with live disease detection overlays.
Bridge telemetry is stale or disconnected. Check the bridge uplink, RTMP route, and local test app before presenting the live mission view.
Bridge mode is currently local-only and intended for controlled demos.
No recent command.
Annotated Drone Stream
RTMP not tested yet.
Start an RTMP stream to mirror the live airborne scan with YOLO overlays.
Active Model Profile
ModelPotato Disease
Classes3
Input size640×640
Taskdetection
Mission Telemetry
Mission
--
Ground Speed
-- m/s
GPS
--
Failsafe
--
Bridge
Waiting
Reference Guide

Notes, mobile handoff, and API endpoints

This area supports demo flow, client integration, and future Flutter/mobile packaging without changing the backend contract.

1
Open the service
The service is live at noufeldouakh.me. For local demos, run start.bat or Uvicorn and open https://localhost:8000.
2
Lead with image mode
For the clearest story, start with single-image diagnosis before moving to live webcam or drone workflows.
3
Use live mode next
Switch to webcam to show that the same backend can support continuous inference and richer operator feedback.
4
Finish with drone vision
Drone mode turns the project from classifier demo into a future-ready precision agriculture platform.
GET
/health
Returns API health and model loading state.
POST
/predict/image
Uploads an image and returns prediction data with an annotated output URL.
POST
/predict/video
Runs frame sampling over video and returns a rendered MP4 plus class distribution summary.
POST
/predict/frame
Handles single-frame webcam inference for the realtime loop.
POST
/drone/bridge/telemetry
Receives telemetry pushed from the local drone bridge app.
GET
/drone/bridge/commands
Allows the bridge app to poll queued commands and update their execution state.
import requests url = "https://noufeldouakh.me/predict/image" files = {"file": open("leaf.jpg", "rb")} res = requests.post(url, files=files).json() print(res["prediction"]["class"]) print(res["prediction"]["confidence"]) print(res["result_url"])
curl -X POST https://noufeldouakh.me/predict/image \ -F "file=@leaf.jpg"
Service state
Online-ready
FastAPI heartbeat for demos, testing, and local presentation.
Model package
YOLO runtime
Primary plant disease weights and inference pipeline.
Image target
< 1s / image
Fast enough for classroom demos and practical leaf checks.
Live cadence
~0.8s / frame
Adaptive loop for webcam mode and future mobile streaming.
Flight mode
IDLE
Current bridge-reported mission mode.
Battery
--%
Estimated airframe power reserve.
Altitude
--m
Height above takeoff point.
Link quality
--%
Health of the telemetry bridge.
Our Vision

Beyond Just Detection

AgroVision9 combines computer vision with agricultural expertise to provide actionable insights for sustainable farming.

🌽

23+ Plant Types

Broad coverage across Vegetables (Tomato, Pepper, Cucumber), Fruits (Apple, Citrus, Strawberry), and key staples like Corn, Basil, and Beans.

🔍

80 Technical Labels

Advanced YOLOv8 detection for specific agricultural labels, ensuring precise identification of even the most subtle physiological changes.

🧪

30+ Pathogens

Specialized diagnostics for Blights, Viruses, Fungal molds, Mites, and Cankers, with automated severity mapping for every detection.

Ready to protect your harvest?

Download the mobile app today and take AgroVision9 directly into the field.

Download Android APK