Handtracking

{
    'bbox': [x, y, width, height],  # Pixel coordinates
    'score': confidence,              # 0.0 to 1.0
    'label': 'hand'                   # Detection label
}

# Clone repository
git clone https://github.com/victordibia/handtracking.git
cd handtracking

# Install dependencies (TensorFlow 1.x compatible)
pip install tensorflow==1.15.0 opencv-python numpy

# Run hand detection on video
python run.py \
    --input_video your_egocentric.mp4 \
    --output_video output_labeled.mp4 \
    --threshold 0.5  # Confidence threshold

import handtracking

# Initialize detector
detector = handtracking.HandDetector()

# Process webcam stream
detector.detect_from_webcam(
    display=True,
    save_video=False,
    confidence_threshold=0.6
)

<script src="https://cdn.jsdelivr.net/npm/handtrackjs/dist/handtrack.min.js"></script>

<script>
const model = await handTrack.load();
const video = document.getElementById('video');

// Detect hands in video stream
const predictions = await model.detect(video);
predictions.forEach(prediction => {
    console.log(prediction.bbox);  // [x, y, width, height]
    console.log(prediction.score); // Confidence score
});
</script>

# Clone repository
git clone https://github.com/victordibia/handtracking.git
cd handtracking

# Create virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install tensorflow==1.15.0
pip install opencv-python numpy pillow

# Download pre-trained model
# Model will be automatically downloaded on first run

# For web applications
npm install handtrackjs

# Or use directly from CDN

import cv2
from handtracking import HandDetector

# Initialize detector
detector = HandDetector()

# Load video
cap = cv2.VideoCapture('egocentric_video.mp4')

# Get video properties
fps = int(cap.get(cv2.CAP_PROP_FPS))
width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))

# Setup video writer
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
out = cv2.VideoWriter('output_labeled.mp4', fourcc, fps, (width, height))

# Process frames
while cap.isOpened():
    ret, frame = cap.read()
    if not ret:
        break

    # Detect hands
    detections = detector.detect_hands(frame)

    # Draw bounding boxes
    for det in detections:
        x, y, w, h = det['bbox']
        score = det['score']

        # Draw box
        color = (0, 255, 0) if score > 0.7 else (0, 0, 255)
        cv2.rectangle(frame, (x, y), (x+w, y+h), color, 2)

        # Add label
        label = f"Hand: {score:.2f}"
        cv2.putText(frame, label, (x, y-10),
                   cv2.FONT_HERSHEY_SIMPLEX, 0.5, color, 2)

    # Save frame
    out.write(frame)

cap.release()
out.release()

import cv2
import numpy as np
from handtracking import HandDetector

detector = HandDetector()
cap = cv2.VideoCapture('egocentric.mp4')

frame_count = 0
while cap.isOpened():
    ret, frame = cap.read()
    if not ret:
        break

    detections = detector.detect_hands(frame)

    # Extract and save hand regions
    for i, det in enumerate(detections):
        x, y, w, h = det['bbox']

        # Crop hand region
        hand_roi = frame[y:y+h, x:x+w]

        # Save hand image
        if det['score'] > 0.7:  # High confidence only
            cv2.imwrite(f'hand_{frame_count}_{i}.jpg', hand_roi)

    frame_count += 1

cap.release()

from handtracking import HandDetector
import cv2

detector = HandDetector()
cap = cv2.VideoCapture(0)  # Webcam

while True:
    ret, frame = cap.read()
    if not ret:
        break

    detections = detector.detect_hands(frame)

    # Display statistics
    num_hands = len(detections)
    avg_confidence = sum(d['score'] for d in detections) / num_hands if num_hands > 0 else 0

    # Overlay text
    cv2.putText(frame, f"Hands: {num_hands}", (10, 30),
               cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 255, 0), 2)
    cv2.putText(frame, f"Avg Conf: {avg_confidence:.2f}", (10, 70),
               cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 255, 0), 2)

    cv2.imshow('Hand Tracking', frame)

    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

cap.release()
cv2.destroyAllWindows()

// Load model
const modelParams = {
    flipHorizontal: true,
    maxNumBoxes: 2,
    iouThreshold: 0.5,
    scoreThreshold: 0.6,
};

handTrack.load(modelParams).then(model => {
    // Model loaded
    console.log("Model loaded");

    // Detect from video element
    const video = document.getElementById('video');
    const canvas = document.getElementById('canvas');
    const context = canvas.getContext('2d');

    function detectFrame() {
        model.detect(video).then(predictions => {
            // Clear canvas
            context.clearRect(0, 0, canvas.width, canvas.height);

            // Draw video frame
            context.drawImage(video, 0, 0, canvas.width, canvas.height);

            // Draw predictions
            predictions.forEach(prediction => {
                const [x, y, width, height] = prediction.bbox;
                context.strokeStyle = '#00FF00';
                context.lineWidth = 4;
                context.strokeRect(x, y, width, height);

                // Add label
                context.fillStyle = '#00FF00';
                context.fillText(
                    `Hand: ${prediction.score.toFixed(2)}`,
                    x, y - 10
                );
            });

            // Continue detection
            requestAnimationFrame(detectFrame);
        });
    }

    // Start detection
    detectFrame();
});