Skip to content
All projects
Software Development+ AI/ML

AI-Powered Search & Rescue Platform

Flask web application that runs YOLOv8 object detection on drone footage, with ML inference kept in its own module separate from the web layer. Covers the whole path from image preprocessing through inference to persisted results, so the model is usable from a browser rather than only a notebook.

Key result

Notebook → browser— YOLOv8 inference served behind a REST API

01Problem

A detection model that only runs in a notebook is not usable by the people who need it. Turning one into something a search coordinator can actually operate means handling upload, preprocessing, inference, and persistence as a single application.

02Objective

Wrap YOLOv8 object detection in a web application so drone footage can be uploaded, processed, and reviewed from a browser, with results stored rather than discarded after each run.

03Architecture

  1. Web Application
  2. Flask Backend
  3. Image Processing
  4. YOLOv8 Inference
  5. PostgreSQL
From uploaded imagery to persisted detection results

04Technology

  • Python
  • Flask
  • PostgreSQL
  • YOLOv8
  • OpenCV

05Implementation

Key technical decisions

  1. 01Kept ML inference in its own module, separate from the web layer, so the model can change without touching request handling.
  2. 02Exposed detection through REST endpoints rather than coupling it to page rendering.
  3. 03Persisted detection metadata and search history in PostgreSQL so results are reviewable after the fact.
  4. 04Covered the full path from image preprocessing through inference to stored results, rather than stopping at model output.

What was built

  • Flask web application running YOLOv8 object detection on drone footage.
  • REST endpoints handling image upload, detection, and results retrieval.
  • OpenCV image preprocessing ahead of inference.
  • Detection metadata and search history stored in PostgreSQL.

06Data

Imagery in, structured detections out

  • Drone footage uploaded through the application
  • Preprocessed images passed to the YOLOv8 detector
  • Detection metadata persisted to PostgreSQL
  • Search history retained for later retrieval

07Results

YOLOv8

Object detection over drone footage

Modular

Inference separated from the web layer

REST

Upload, detect, retrieve endpoints

PostgreSQL

Detection metadata and history

10Links

GitHub