How tssanjayy/handgesture works
This is a real-time hand-gesture recognition project with a data-to-model workflow and two live recognition paths: the documented Random Forest demo and a separate rule-based recognizer. The graph covers session capture, offline feature preparation and evaluation, plus live prediction and visualization. Some processing is implemented inline in the demo scripts; the README documents the main capabilities, while the sampled source provides the cited runtime relationships.
Main entry points: live_demo_ui.py Live ML interface · live_demo.py Live ML demo · data_collection.py Session collector · model_engineered.joblib Engineered model
Components 16
Data capture
- Session collector
data_collection.py - Session CSVs
data
Feature preparation
- Feature extraction
features.py - Feature datasets
features
Modeling and evaluation
- Final evaluation
final_evaluation.py - Training evaluation
train_and_evaluate.py - Raw model
model_raw.joblib - Engineered model
model_engineered.joblib
Live recognition
- Live ML interface
live_demo_ui.py - Live ML demo
live_demo.py - Rule-based recognizer
gesture_recognition.py - Hand tracker
hand_tracking.py
Other components
- User
- Webcam
- MediaPipe Hands
- OpenCV
Connections 28
- User → Session collector: labels samples
data_collection.py - Webcam → Session collector: provides frames
data_collection.py - Session collector → MediaPipe Hands: detects landmarks
data_collection.py - Session collector → Session CSVs: writes samples
data_collection.py - Session CSVs → Feature extraction: loads sessions
features.py - Feature extraction → Feature datasets: writes features
features.py - Feature datasets → Final evaluation: loads datasets
final_evaluation.py - Final evaluation → Raw model: saves model
final_evaluation.py - Final evaluation → Engineered model: saves model
final_evaluation.py - Feature datasets → Training evaluation: loads datasets
train_and_evaluate.py - Training evaluation → Raw model: saves model
train_and_evaluate.py - Training evaluation → Engineered model: saves model
train_and_evaluate.py - User → Live ML interface: selects model
live_demo_ui.py - Webcam → Live ML interface: provides frames
live_demo_ui.py - Live ML interface → MediaPipe Hands: detects landmarks
live_demo_ui.py - Live ML interface → OpenCV: captures and renders
live_demo_ui.py - Raw model → Live ML interface: loads selected model
live_demo_ui.py - Engineered model → Live ML interface: loads selected model
live_demo_ui.py - Webcam → Live ML demo: provides frames
live_demo.py - Live ML demo → MediaPipe Hands: detects landmarks
live_demo.py - Live ML demo → Engineered model: loads and predicts
live_demo.py - Live ML demo → OpenCV: captures and renders
live_demo.py - Webcam → Rule-based recognizer: provides frames
gesture_recognition.py - Rule-based recognizer → MediaPipe Hands: detects landmarks
gesture_recognition.py - Rule-based recognizer → OpenCV: captures and renders
gesture_recognition.py - Webcam → Hand tracker: provides frames
hand_tracking.py - Hand tracker → MediaPipe Hands: detects landmarks
hand_tracking.py - Hand tracker → OpenCV: captures and renders
hand_tracking.py