BSD 3-Clause "New" or "Revised" License

项目介绍

WangWilly/gaze-correction-cam 是一个托管在 GitHub 上的开源项目,主要使用 Python 编写,采用 BSD 3-Clause "New" or "Revised" License 许可证,目前获得 114 个 Star、21 次 Fork。

仓库全名
WangWilly/gaze-correction-cam
开源许可证
BSD 3-Clause "New" or "Revised" License
Star 数
114
Fork 数
21
未关闭 Issue
5

README 摘要

Gaze Correction Camera Make natural eye contact on every video call. Gaze Correction Camera is a macOS virtual camera extension that automatically adjusts your eye direction in real-time — so you always look directly at the people you're talking to, even when you're reading notes or looking at your own video tile. Demo Download ⬇️ Download the macOS App Install the .app, grant camera access, and you're ready to go…

完整内容见 GitHub 原始 README 或本页 。

常见问题

gaze-correction-cam 用什么语言开发?

主要使用 Python 编写,采用 BSD 3-Clause "New" or "Revised" License 许可证,目前有 114 个 Star。

gaze-correction-cam 可以免费商用吗?

仓库采用 BSD 3-Clause "New" or "Revised" License 许可证,可免费使用与修改,商业使用请遵守许可证条款(例如保留版权声明与许可声明)。

如何获取 gaze-correction-cam 的源码?

可以直接克隆仓库:git clone https://github.com/WangWilly/gaze-correction-cam.git,默认分支为 master;安装与运行步骤见仓库 README。

gaze-correction-cam 最近还在维护吗?

仓库最近一次更新为 3月12日,当前有 5 个未关闭 Issue;Star 数为 114,可作为社区活跃度的参考。

关于

GitHub 原始描述(英文)

The Gaze Correction Camera project is an advanced real-time gaze correction system designed to enhance video communication by improving eye contact. Leveraging state-of-the-art computer vision and deep learning techniques, this system dynamically adjusts the user's eye gaze direction during live video calls.

描述取自 GitHub 官方 API,保持仓库作者书写的原文。

语言占比

Python

Release

贡献者

README

Gaze Correction Camera

Make natural eye contact on every video call.

Gaze Correction Camera is a macOS virtual camera extension that automatically adjusts your eye direction in real-time — so you always look directly at the people you’re talking to, even when you’re reading notes or looking at your own video tile.

Demo

Download

⬇️ Download the macOS App

Install the .app, grant camera access, and you’re ready to go.

Requirements

  • macOS 14 (Sonoma) or later
  • A connected webcam or built-in camera
  • Camera access permission granted to the app

Getting Started

  1. Download and open the app from the link above.
  2. Grant camera access when prompted by macOS.
  3. Select your camera from the dropdown in the top-right corner of the app window.
  4. Enable gaze correction using the toggle in the settings panel.
  5. Set this app as your camera in Zoom, Teams, FaceTime, or any video app — look for “Gaze Correction Camera” in the camera device list.

Controls

Key Action
g Toggle gaze correction on / off
c Toggle calibration panel
q Quit

Calibration

Press c to open the calibration panel and fine-tune the correction for your setup:

Control Action
↑ ↓ ← → Adjust camera position up/down/left/right
+ / - Adjust camera distance (closer/further)
[ / ] Adjust focal length
r Reset all values to default

Tip: Start with the defaults. Only calibrate if the gaze correction looks off for your specific desk setup.


Advanced: Build from Source / Python CLI

If you want to run from source or use the Python CLI directly:

Prerequisites

Install

brew install pkg-config cmake
poetry install

Download model weights

Download from GitHub Releases and place in the correct directories:

  • lm_feat/shape_predictor_68_face_landmarks.dat
  • weights/warping_model/flx/12/L/ and .../R/ — checkpoint + weight files
  • (Optional) models/face_landmarker.task — for MediaPipe backend

Run

# Default (dlib backend)
poetry run python bin_single_window.py

# MediaPipe backend
poetry run python bin_single_window.py --backend mediapipe

# Specific camera
poetry run python bin_single_window.py --camera 1

How It Works

  1. Captures your webcam feed in real-time
  2. Detects your face and eye positions using computer vision
  3. Calculates where your eyes need to point to look at the camera
  4. Applies a learned neural network warp to redirect your gaze
  5. Outputs the corrected video as a virtual camera device

Looking for architecture details or module documentation? See docs/architecture.md.

评论

暂无评论