README

Video AI Object Detector - Offline Audit Tool

A high-performance, multithreaded offline video auditing tool designed for automated object detection and visual verification. Built using Lazarus (Free Pascal) for a responsive GUI and YOLOv8 (Python) for computer vision analysis, the application enables frame-accurate log tracking, evidence snapshot generation, and seamless video navigation.

Key Features

• Offline & Private: Analyzes videos locally without requiring an internet connection or external API calls.
• Non-Blocking Multithreading: Powered by Python daemon threading and thread-safe data queues in Lazarus to keep the UI fully responsive during heavy inference.
• Real-Time Detection Logs: Captures detected objects along with exact video timestamps and allows live keyword filtering.
• Frame-Accurate Video Scrubbing: Double-clicking any log entry automatically synchronizes PasLibVlcPlayer to the exact millisecond and displays the annotated evidence snapshot.
• Custom Inference Controls: Configurable target FPS (frame skipping) and confidence thresholds to balance speed and accuracy.
• Automated Reporting: Generates text reports (result.txt) and scaled detection snapshots automatically.

Tech Stack

• GUI / Frontend: Lazarus IDE (Free Pascal Compiler)
• Media Engine: PasLibVlcPlayer (libVLC integration)
• AI Engine: YOLOv8 via Python 3.12 (Embedded Python Engine / Python4Lazarus)
• Computer Vision: OpenCV (cv2) & PyTorch

System Requirements & Directory Structure

Ensure the output executable directory includes the following dependencies:

├── your_app.exe
├── engine_ai.py
├── yolov8n.pt
├── crash_debug.txt (auto-generated)
├── python_embed/           # Embedded Python 3.12 runtime
│   ├── python312.dll
│   └── ...
├── vlc_libs/               # Portable libVLC binaries and plugins
│   ├── libvlc.dll
│   ├── libvlccore.dll
│   └── plugins/
└── temp_snapshots/         # Auto-managed folder for snapshots and logs

☕ Support the Project

If you find AVIDAI helpful and want to support its ongoing development, consider buying me a coffee or sending a tip. Any support is deeply appreciated!

💡 Your support keeps the momentum going!
Every contribution directly fuels my passion, energy, and motivation to continuously build, maintain, and release even more useful open-source desktop applications for the developer community.