Pipeline Explorer // Sign to Speech
PAUSE
STEP
TOGGLE VIEW
RESET
01 // IMAGE_ACQUISITION
02 // LANDMARK_MAP
03 // SPEECH_OUT
TENSOR
VEC
FEED_SOURCE: DEV/VIDEO0
ROI: DETECTED
MASK_VIEW
SWITCH_VIEW
SKELETON_MODEL [21_PTS]
NODE_STATUS: IDLE
CLASSIFICATION_OUT
A
p: 0.02
W
p: 0.10
Y
p: 0.88
SMOOTHING_TIMELINE (last 10 frames)
SYNTHESIZE: "Y"
Captures 60FPS camera frames,
isolating hand ROI from background.
Extracts 21 distinct 3D landmarks
representing joints and finger positions.
Smooths classifier timeline &
triggers Text-to-Speech utterance.