face-analysis-kit 0.1.0

A package for analyzing faces in images to detect eye state, gaze direction, and facial expressions

Face Analysis

A package for analyzing faces in images to detect eye state, gaze direction, and facial expressions.

Features

• Gaze Detection: Determine gaze direction using ResNet models
• Eye State Classification: Detect whether eyes are open or closed
• Emotion Recognition: Identify facial expressions and emotions

How to install:

1. Create a virtual environment and install dependencies:

make install

This will:

• Install all dependencies using uv
• Set up pre-commit hooks for code quality

2. Activate the environment:

source .venv/bin/activate

Usage Examples

Command Line Interface

The package provides several command-line tools for analyzing faces in images:

1. Analyze a single image:

bash scripts/run_analysis.sh single input/test_1.png

2. Analyze all images in a folder:

bash scripts/run_analysis.sh folder input/

3. Select the best image from a folder:

bash scripts/run_analysis.sh best input/

Output will be saved to output/<command>/ directory.

Python API

Gaze Detection

from face_analysis.gazes import Pipeline as GazesPipeline
from face_analysis.gazes import render as GazesRender

gaze_pipeline = GazesPipeline(
    arch='ResNet50',  # Options: "ResNet18", "ResNet34", "ResNet101", "ResNet152"
    detector="retinaface",  # Options: "mtcnn"
    device="cuda",  # or "cpu"
)

img_in = cv2.imread("input/test_1.png")
results = gaze_pipeline.step(img_in)
img_out = GazesRender(img_in, results)

Eye State Detection

from face_analysis.eyes import Pipeline as EyesPipeline
from face_analysis.eyes import render as eyes_render

eye_pipeline = EyesPipeline(
    detector="retinaface", # or "dlib"
    device="cpu", # or "cuda"
)

img_in = cv2.imread(img_path)
results = eye_pipeline.step(img_in)
img_out = eyes_render(img_in, results)

Emotion Detection

from face_analysis.emotions import Pipeline as EmotionsPipeline
from face_analysis.emotions import render as emotions_render

emotion_pipeline = EmotionsPipeline(
    detector= "retinaface", # or "mtcnn", or "cascade"
    device= "cpu",
)

img_in = cv2.imread(img_path)
results = emotion_pipeline.step(img_in)
img_out = emotions_render(img_in, results)

Repo Structure

project_root/
├── data/       
├── input/
├── output/
├── scripts/
├── face_analysis/                 
│
├── LICENSE.txt
├── pyproject.toml
├── README.md
├── requirements.txt
├── uv.lock
└── Makefile