Real-Time Static Hand Gesture Recognition

Your Full Name: Rohan Patil

---

Overview

This project implements a real-time Python application that uses a webcam to detect and recognize a set of static hand gestures. It captures a live video feed, detects and tracks a single hand, extracts landmark-based features, and classifies the current hand pose into one of the supported gestures — all with on-screen feedback and optional recording.

---

Gesture Vocabulary

The application recognizes four static gestures:

1. Open Palm
2. Fist
3. Peace Sign (V-sign)
4. Thumbs Up

A detected gesture’s name is displayed on the output video window in real time.

---

Technology Choice & Justification

• MediaPipe Hands — for hand detection and 21 landmark extraction. Fast and CPU-friendly.
• OpenCV — for webcam capture, drawing overlays, and recording.
• NumPy — for numeric operations (angles, distances, normalization).
• scikit-learn + joblib — optional, for training lightweight classifiers.

Why this stack?
MediaPipe + OpenCV ensures robust, low-latency detection suitable for real-time use while keeping the app lightweight.

---

How the System Works

1. Capture video frames with OpenCV.
2. Detect hand and extract landmarks with MediaPipe.
3. Normalize landmarks to scale/position.
4. Extract features (angles, distances, orientation).
5. Apply rules-based classifier:
   • Open Palm: all fingers extended.
   • Fist: all fingers folded.
   • Peace: index & middle extended, others folded.
   • Thumbs Up: four fingers folded, thumb extended upwards.
6. Smooth predictions with a sliding window + debounce.
7. Display gesture name and FPS on screen.
8. Optionally record annotated video.

---

Installation

Run each installation command separately:

pip install opencv-python
pip install mediapipe
pip install numpy
pip install scikit-learn
pip install joblib

⚠️ If mediapipe fails:

Make sure Python version is 3.8–3.11

Upgrade pip:

python -m pip install --upgrade pip

---

Running the App

python app.py

Controls

• q → quit
• c → calibrate (show open palm for 2s)
• d → toggle data capture

Recordings are saved in the demo/ folder.

---

Project Structure

hand-gestures/
├─ src/
│  ├─ app.py                 # main loop, calibration, recording
│  ├─ features.py            # normalization & feature extraction
│  ├─ gestures.py            # rules-based classifier with smoothing & calibration
│  └─ utils.py               # drawing helpers and FPS meter
├─ demo/
│  └─ recording_YYYYMMDD_HHMMSS.avi
├─ captured_data.csv         # optional: features + labels
└─ README.md

---

Demo

A demonstration video is included in the demo/ folder:

demo/demo.mp4

If you run the app, an annotated .avi file will also be saved automatically.

---

Performance

• Runs in real-time (>25 FPS on CPU).
• Adjustable smoothing to reduce flicker.
• Works on standard laptop webcam input.

---

Troubleshooting

• mediapipe not installing → check Python version (3.8–3.11 only).
• Low FPS → reduce webcam resolution or set model_complexity=0.
• Flickering gestures → increase smoothing window/debounce values.

---

Contact

For improvements or questions:
[email protected]

Code Quality

• Code is modular and organized by functionality.
• Inline comments explain complex or non-obvious logic.
• Functions are named descriptively for readability.

---

Requirements

• Python 3.7+
• OpenCV
• MediaPipe

Install dependencies with:

pip install -r requirements.txt

"# Real-Time-Static-Hand-Gesture-Recognition"