How liz003-ziz1010/activity-recognition works
The repository contains two related but distinct gesture projects: a Flask app for capturing labeled hand images and predicting gestures, and a webcam-motion toolkit for recording, preparing, and classifying action videos. The toolkit also includes standalone landmark recording and pose-processing workflows. The map follows the sampled Flask, training, dataset, inference, motion-detection, and pose-processing code; wiring for other files is not inferred from their names alone.
Main entry points: app.py Flask routes · train_classifier.py Classifier training · inference_classifier.py Video inference · train.py Image training
Components 18
Hand Gesture Web App
- Flask routes
hand_gesture_app/app.py - Captured images
- Image training
hand_gesture_app/train.py - Hand CNN
hand_gesture_app/model.py
Video Action Classification
- Motion detector
webcam-motion2.0/webcam-motion.py - Action clips
webcam-motion2.0/data - Video dataset
webcam-motion2.0/video_dataset.py - Classifier training
webcam-motion2.0/train_classifier.py - Video classifier
webcam-motion2.0/action_classifier.py - Saved weights
webcam-motion2.0/action_classifier_model.pth - Video inference
webcam-motion2.0/inference_classifier.py
Pose And Landmark Tools
- Landmark recorder
webcam-motion2.0/gesture.py - Gesture recorder and predictor
webcam-motion2.0/gestures_saved.py - Pose extraction
webcam-motion2.0/pose_extractor.py - Pose and skeleton processing
webcam-motion2.0/pose_segment_and_skeleton.py
Other components
- Browser user
- Webcam
- Motion-tool user
Connections 23
- Browser user → Flask routes: uses routes
- Flask routes → Webcam: reads frames
- Flask routes → Captured images: saves captures
- Flask routes → Image training: starts training
- Image training → Captured images: loads images
- Image training → Hand CNN: trains
- Flask routes → Hand CNN: predicts frame
- Motion-tool user → Motion detector: starts capture
- Motion detector → Webcam: reads frames
- Motion detector → Action clips: records clips
- Classifier training → Video dataset: loads samples
- Video dataset → Action clips: reads clips
- Classifier training → Video classifier: trains
- Classifier training → Saved weights: saves weights
- Video inference → Video classifier: runs classification
- Video inference → Video dataset: infers classes
- Video inference → Saved weights: loads weights
- Motion-tool user → Landmark recorder: records or replays
- Landmark recorder → Webcam: reads frames
- Motion-tool user → Gesture recorder and predictor: runs gesture tool
- Gesture recorder and predictor → Webcam: reads frames
- Motion-tool user → Pose extraction: processes video
- Motion-tool user → Pose and skeleton processing: processes video