Interactive Web Showcase Notice
This page is an interactive website showcasing the entire DEXTRA system architecture, 21-landmark poses, and voice command engine โ it is a web-based presentation view, not the desktop backend production daemon.
Real-time MediaPipe skeletal joint visualization & PyAutoGUI gesture heuristic decoder
Click a pose preset on the right to animate 21 joint nodes and view live telemetry math.
Comprehensive touchless input taxonomy for seamless desktop navigation
Powered by local Hugging Face Transformer Whisper-tiny with 'DEXTRA' Wakeword Monitor
Click any command chip below to simulate background microphone voice activation and view decoded STT transcription.
Concurrent daemon threads for high-frame-rate tracking and instantaneous voice execution
Captures 30+ FPS BGR frames from internal camera index 0 or USB external mic/webcam.
python/camera.py
Extracts 21 3D coordinates & calculates Euclidean distances for finger posture heuristics.
python/gesture_engine.py
Applies exponential moving average buffer to eliminate jitter before PyAutoGUI dispatch.
python/gesture_mapping.py
Background sounddevice listener triggering local Hugging Face Whisper model upon 'DEXTRA' wakeword.
python/voice_engine.py
Live interactive parameter tuning mirroring dextra_settings.json schema
Move your cursor inside the container below to experience how EMA smoothing softens jitter in real-time.
{
"camera_index": 0,
"cursor_smoothing": 5,
"click_threshold_px": 30,
"drag_hold_time_sec": 0.6,
"double_click_time_sec": 0.35,
"cooldown_period_sec": 0.3,
"wakeword": "DEXTRA"
}
Derived live from DEXTRA_Project_Tracker.xlsx management sheet
Conceived initial system concept, architectural vision & gesture-driven paradigm.
MediaPipe 21 Landmarks, OpenCV Camera Pipeline, PyAutoGUI Mouse Events.
Sounddevice Mic Engine, Hugging Face Whisper ASR, Shortcut Key Mapper.
FastAPI REST Endpoints, WebSockets Image Streamer, Thread Safety Locks.
React SPA Settings GUI, HTML5 Canvas Trainer Streamer, Design System.
| Bug ID | Target Module | Severity | Description | Owner | Status | Resolution Details |
|---|---|---|---|---|---|---|
B01 |
python/camera.py |
Medium | VideoCapture camera locking on server restart causing frame capture freeze. | Salman | Closed | Added try-finally cleanup blocks and explicit camera.release() calls. |
B02 |
python/gui_server.py |
High | FastAPI WebSocket stream disconnect error on /trainer/stream client closure. | Muthamil | Closed | Handled WebSocketDisconnect gracefully & added client connection manager. |
B03 |
python/gesture_engine.py |
Medium | MediaPipe clearcut telemetry upload connection errors (Status_ConnectFailed: 12029). | Salman | Closed | Disabled Clearcut telemetry background uploads via environment configuration. |
B04 |
python/gesture_mapping.py |
High | Mouse cursor jitter and accidental double-clicks during fast movement. | Salman | Closed | Implemented EMA smoothing filter (alpha=0.35) and click cooldown timer. |
B05 |
python/voice_wakeword.py |
Medium | High CPU utilization during continuous audio stream buffer recording. | Logesh | Closed | Optimized sounddevice blocksize to 1024 samples with ring buffer queue. |
B06 |
Git Repository |
Low | Default tracking branch misalignment on GitHub remote. | Godfrey | Closed | Updated local remote HEAD to dextra branch and pruned obsolete remotes. |