Quartz 4

setup

2 min read

Setup Guide

Build & Run from Source

Prerequisites

  • NVIDIA GPU Required for ZED SDK, InsightFace, and GFPGAN inference.
  • CUDA Must have CUDA 11.x or 12.x installed matching your ZED SDK version.
  • ZED Camera ZED 2 or ZED 2i connected via USB 3.0.
  • ZED SDK Install the latest ZED SDK for your OS.
  • MongoDB Must be installed and running locally on port 27017.
  • Python 3.10 or 3.12 recommended.

1. Installation

Clone the repository:

git clone https://github.com/yourusername/fnr-facial-net-recognizer.git
cd fnr-facial-net-recognizer

Set up Python Environment:

Bash

python -m venv venv
source venv/bin/activate  # On Windows use: venv\Scripts\activate

Install Dependencies:

Bash

# Install core libraries
pip install opencv-python numpy pymongo flask streamlit
 
# Install AI engines (GPU versions)
pip install insightface onnxruntime-gpu
pip install gfpgan basicsr
 
# Install ZED Python API (if not installed by SDK)
# Navigate to: "C:\Program Files (x86)\ZED SDK\python" and run:
python get_python_api.py
  1. Download Model Weights The system requires the GFPGAN pre-trained model for super resolution.

Bash

# Download GFPGANv1.3.pth to the root directory
wget https://github.com/TencentARC/GFPGAN/releases/download/v1.3.0/GFPGANv1.3.pth

(On Windows, you can manually download this file from the link above and place it in the project folder)

  1. Start Database Ensure your MongoDB instance is running. If you have Docker installed, this is the easiest way:

Bash

docker run -d -p 27017:27017 --name face-db mongo:latest
  1. Run the System You will need two separate terminal windows.

Terminal 1: The Backend (Camera & AI Engine) This script captures video, runs recognition, and hosts the video stream server.

Bash

python main.py
# Wait until you see: "System Running!"
# Note the IP address displayed: e.g., [http://192.168.1.50:5001/video_feed](http://192.168.1.50:5001/video_feed)

Terminal 2: The Dashboard (Frontend) Important: Open dashboard.py and ensure STREAM_URL matches the IP from Terminal 1.

Bash

streamlit run dashboard.py

Open your browser to the URL shown (usually http://localhost:8501).

Tailscale Usage Use Tailscale IP for remote access to the Flask stream.

Update STREAM_URL in the dashboard script with your Tailscale IP (e.g., http://100.x.x.x:5001/video_feed).

Tailscale allows secure connection between the camera server and the dashboard without exposing ports to the public internet.

System Specifications

Minimum

  • CPU Intel i7 or equivalent
  • RAM 16GB or more
  • GPU NVIDIA RTX 20 series or newer
  • Disk 10GB free space

Recommended

  • CPU Intel i9 or equivalent
  • RAM 32GB or more
  • GPU NVIDIA RTX 3080 or newer for best performance
  • VRAM 8GB or more

See Also Project Overview

Development Iterations

Key Imports and Technologies

Graph View