Skip to content

Latest commit

 

History

History
182 lines (132 loc) · 4.19 KB

README.md

File metadata and controls

182 lines (132 loc) · 4.19 KB

VisionixAI

Zone-Based Computer Vision Automation
Smart actions through visual presence detection—no external sensors or hardware dependencies required.

VisionixAI is a modular computer vision platform that detects human presence within predefined spatial zones and orchestrates automated responses. By leveraging high-performance visual processing, VisionixAI delivers seamless, scalable automation for smart environments—from homes to enterprise settings.

Table of Contents

System Overview

Any physical space is divided into virtual grid zones. When a zone remains unoccupied for a configurable duration, VisionixAI emits control signals to power off connected devices. Upon detecting a person re-entering the zone, the system reverses the automation. All interactions rely solely on visual input.

Key features:

  • Zone Partitioning: Virtual grid overlay on frames.
  • Presence Detection: Combines object detection (YOLOv8) and tracking (MediaPipe).
  • Noise Reduction: Temporal smoothing to prevent false triggers.
  • Extensible Output: MQTT, GPIO, WebSockets, or cloud relays.

Architecture Diagram

flowchart TD
  subgraph EntryPoint
    A1[CLI Tool]
  end

  subgraph Input
    B1[Camera Stream Handler]
    B2[Static Video Loader]
  end

  subgraph Preprocessing
    C1[Frame Normalizer]
    C2[Zone Grid Mapper]
  end

  subgraph Detection_Engine
    D1[YOLOv8 Detector]
    D2[MediaPipe Tracker]
    D3[Frame-wise Inference]
  end

  subgraph Postprocessing
    E1[Zone Occupancy Logic]
    E2[Temporal Smoothing]
  end

  subgraph Output_Layer
    F1[Signal Dispatcher]
    F2[MQTT Emitter]
    F3[GPIO / Relay Trigger]
  end

  subgraph API_Layer
    G1[FastAPI Interface]
    G2[RESTful Endpoints]
    G3[Web Dashboard]
  end

  subgraph Data_Storage
    H1[Local State]
    H2[SQLite or JSON]
  end

  subgraph DevOps_and_Tooling
    I1[Typer CLI]
    I2[Rich Terminal UI]
    I3[Logging Module]
  end

  A1 --> B1
  A1 --> B2

  B1 --> C1
  B2 --> C1

  C1 --> C2
  C2 --> D1
  C2 --> D2

  D1 --> D3
  D2 --> D3

  D3 --> E1
  E1 --> E2

  E2 --> F1
  F1 --> F2
  F1 --> F3

  A1 --> I1
  A1 --> I2
  A1 --> I3

  A1 --> G1
  G1 --> G2
  G2 --> G3

  E2 --> H1
  H1 --> H2
Loading

Components

1. CLI Tool

  • Entry point for live camera or video input.
  • Built with Typer and Rich for UX.

2. Input Handlers

  • Camera Stream Handler: Captures real-time webcam feed.
  • Static Video Loader: Processes pre-recorded footage.

3. Frame Preprocessing

  • Frame Normalizer: Resizes and standardizes frames.
  • Zone Grid Mapper: Overlays a configurable grid for zone partitioning.

4. Detection Engine

  • YOLOv8 Detector: High-performance object detection.
  • MediaPipe Tracker: Robust human tracking across frames.
  • Integrated inference pipeline for per-zone presence detection.

5. Postprocessing

  • Zone Occupancy Logic: Determines zone transitions.
  • Temporal Smoothing: Aggregates results to reduce noise.

6. Signal Dispatcher

  • Emits automation signals via:
    • MQTT brokers
    • GPIO pins / Relay modules
    • Extendable for WebSockets or cloud relays

7. (Optional) API Layer

  • Planned FastAPI interface for external integration and dashboards.

8. Data Storage

  • Local State: In-memory state persistence.
  • Storage Backend: SQLite or JSON for logging and replay.

9. DevOps & Tooling

  • CI/CD: GitHub Actions workflows.
  • Logging: Structured logs for debugging.

Technologies

  • Language: Python 3.11+
  • Computer Vision: OpenCV, YOLOv8 (Ultralytics), MediaPipe
  • CLI & UX: Typer, Rich
  • API: FastAPI (planned)
  • Signaling: MQTT, GPIO

Status

  • CLI Pipeline: Functional and modular.
  • Detection Engine: YOLOv8 + MediaPipe integrated and tested.
  • Signal Dispatch: MQTT signaling verified.
  • API & Dashboard: In development.

License

This project is licensed under the MIT License.