Face Tracking Project

A high-performance system for real-time face detection and tracking. This repository contains both the Python-based training and the C++ deployment.

Project Structure

train/: Model training and optimization using Python and uv.

deploy/: High-performance C++ inference

model/: Storage for exported TorchScript model weights.

data/: Local video files and sample images for testing.

doc/: Documentations

Technical Specifications

Deep Learning: PyTorch (Training) / LibTorch (Deployment).

Getting Started Python Training (train/) Ensure uv is installed on your system.

Navigate to the directory and sync dependencies:

cd train
uv sync

Run the export script to generate the TorchScript model:

uv run python export.py

C++ Deployment (deploy/) Configure OpenCV and LibTorch paths in CMakeLists.txt.

Use CLion or Visual Studio with the MSVC (x64) toolchain.

Build the project:

cd deploy

Using CMake to configure and build

cmake -B cmake-build-debug -G Ninja
cmake --build cmake-build-debug

Run the executable to start camera-based tracking.