How me-sarvar/opencv works
Project type and purpose
• This is a local, Python-based gesture‐recognition application. It uses OpenCV and MediaPipe to capture hand landmarks, collects gesture data, trains an LSTM model with TensorFlow/Keras, and then runs a real-time inference loop to recognize one of five predefined gestures via webcam.
File structure
• Top-level files
• .gitignore – ignore rules
• README.md – project overview
• requirements.txt – Python dependencies
• python/ directory: contains the three main scripts
• data_collect.py – data‐acquisition component
• model_train.py – model‐training pipeline
• main.py / new_main.py – real‐time inference loop
• Implicit output directories (created at runtime):
• data_set/ – raw gesture feature sequences
• models/ – saved LSTM model artifacts
README for additional insights
• Describes five target words (words array) mapped to hand gestures
• Outlines three phases: data collection, model training, and real-time recognition
• Highlights ability to customize model architecture and gestures
Detail
• Include:
• Exact script names and paths
• Data formats (e.g., hand‐landmark CSV or NumPy arrays)
• Model file format (e.g., .h5)