📷 Camera Assistant

A live computer-vision viewer built with Python 3, OpenCV, and Tkinter — powered by your phone camera via scrcpy + v4l2loopback.

Demo pipeline

Phone (USB/ADB) → scrcpy v4l2-sink → /dev/video0 → OpenCV → Tkinter canvas

Features

Toggle	Detection
👤 Face	Blue bounding box + "Face" label
👁 Eye	Amber boxes inside detected faces
😊 Smile	Yellow highlight on smiles
🤚 Hand	OpenCV skin contour + convexity defects + finger count bar
🧍 Body	Purple boxes (full + upper body)
⚡ Edge	Canny edge overlay (teal tint)

• FPS counter displayed at top-left
• Finger counting via convexity defects — YCrCb skin segmentation + contour analysis
• Smart camera scan — only checks existing /dev/video* devices
• Dark theme — Catppuccin Mocha inspired, glassmorphism aesthetic

Requirements

# Arch / pacman
sudo pacman -S python-opencv python-pillow python-numpy v4l2loopback-dkms

# scrcpy (phone camera pipe)
sudo pacman -S scrcpy android-tools

# Load v4l2loopback
sudo modprobe v4l2loopback

Package	Why
OpenCV	Face/Haar, YCrCb skin mask, contour detection, all CV
tkinter	Built-in GUI
pillow	Frame → Tkinter canvas

⚠️ No external ML libraries needed — hand detection uses pure OpenCV (YCrCb colour-space skin segmentation + convexity-defect finger counting).

You also need v4l2loopback-dkms built for your kernel (pacman handles this).

Usage

1. Add yourself to the video group (one-time):

sudo usermod -a -G video $USER
# Then log out and back in

2. Start the phone camera stream (keep this terminal open):

scrcpy --v4l2-sink=/dev/video0 --video-source=camera \
       --camera-id=0 --camera-size=1920x1080 \
       --camera-fps=30 --no-window

3. Launch the app (new terminal):

python3 camera_assistant.py

Select Camera 0 from the dropdown → click ▶ Start.

Toggle individual detections on/off while the stream is live.

Controls

Action	Button
▶ Start / ■ Stop	Bottom-left
🔄 Rescan	Re-scan camera devices
Face / Eye / Smile / Hand / Body / Edge	Checkboxes — toggle live

Notes

• Built for Nothing A142P (Android 16) but works with any phone camera accessible via ADB.
• The v4l2loopback kernel module creates a virtual video device that scrcpy writes to.
• Hand detection uses skin-colour HSV masking + convexity defects for finger counting.