Autonomous Drone Vision

⚠️ Note: In the current version, the labels for car, truck, and motor are swapped — detections may show “truck” when it’s actually a car, or “motor” when it’s a truck.

Drone Vision: Real-Time Object Detection
An AI-powered web app that detects vehicles (cars, trucks, buses, motorcycles, and people) from aerial drone images in real time.
Built with FastAPI, YOLOv8, and Google Cloud Run, it automatically runs inference in the cloud and visualizes results directly on the webpage.


A computer-vision platform that processes aerial imagery from drones to automatically detect and classify vehicles, people, and infrastructure using YOLOv8 deep learning models.While built as a vision research prototype, it demonstrates how AI can support data-driven risk assessment, asset monitoring, and exposure management — similar to how insurers or reinsurers analyze satellite or drone data after natural disasters.

Applications to finance & risk:
📊 Risk Exposure Analysis: Detects vehicle density or human presence in industrial zones, relevant for property and casualty (P&C) or motor portfolio risk profiling.
🛰️ Catastrophe Modeling Support: Drone-based imagery can help estimate damage or exposure after floods, storms, or fires — similar to PartnerRe’s or Swiss Re’s use of remote sensing for loss estimation.
⚙️ Automation & Cost Efficiency: Demonstrates how AI pipelines (FastAPI + Cloud Run) can scale automatically, similar to automated claim triage or underwriting systems.
🔍 Data Integrity & Model Governance: Includes model versioning, confidence thresholds, and reproducibility — aligning with actuarial validation principles in Solvency II model governance.
Tech stack:

Ultralytics YOLOv8 — custom-trained model for VisDrone dataset⚙️ FastAPI + Uvicorn — inference microservice with CORS + REST API☁️ Google Cloud Run — serverless GPU-ready deployment🌐 Webflow + JavaScript — interactive frontend integration

Features:
Upload drone images and get instant bounding boxes & labelsScalable backend with cloud auto-deploymentSupports retraining and model versioningOpenAPI documentation for developers

Drone Vision Demo




Webflow
Brand Designer
2025
🚁 Drone Vision: Real-Time Object DetectionAn AI-powered web app that detects vehicles (cars, trucks, buses, motorcycles, and people) from aerial drone images in real time.
Built with FastAPI, YOLOv8, and Google Cloud Run, it automatically runs inference in the cloud and visualizes results directly on the webpage.

Example prediction, current model detects objects mostly correctly, but labels for car, truck, and motor are still swapped.

Want to get in touch?
Drop me a line!

I’d love to hear from you — whether it’s about a project, an opportunity, or just to connect.

NAME
EMAIL ADDRESS
FORM LABEL
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Anar Assistant
Ask about projects, CV, skills & more.
🇬🇧 English • 🇩🇪 Deutsch • 🇲🇳 Монгол