How kumarvivek9088/signlanguagedetectionusingml works
This repository implements a webcam-based sign-language recognizer and the scripts to prepare and train its classifier. The user-facing workflow captures hand frames, extracts landmarks into 30-frame sequences, classifies live sequences, and displays the recognized sign. Separate scripts collect labeled images and convert them into landmark data. The README offers no further capability detail; the tree and sampled sources support this workflow, although the live app references a model-architecture file not present in the tree.
Main entry points: app.py Live recognition · data.py Landmark preparation · function.py Hand processing · collectdata.py Image capture
Components 12
Capture and features
- Image capture
collectdata.py - Labeled images
- Landmark preparation
data.py - Hand processing
function.py - Sequence data
MP_Data
Training
- Model training
trainmodel.py - Trained weights
model.h5
Live recognition
- Live recognition
app.py - Sign prediction
app.py - On-screen output
app.py
Other components
- Operator
- Webcam
Connections 17
- Operator → Image capture: labels frames
- Webcam → Image capture: supplies frames
- Image capture → Labeled images: writes images
- Labeled images → Landmark preparation: provides images
- Landmark preparation → Hand processing: calls
- Hand processing → Landmark preparation: returns landmarks
- Landmark preparation → Sequence data: writes sequences
- Sequence data → Model training: loads sequences
- Model training → Trained weights: saves weights
- Operator → Live recognition: starts recognition
- Webcam → Live recognition: supplies frames
- Live recognition → Hand processing: calls
- Hand processing → Live recognition: returns landmarks
- Trained weights → Live recognition: loads weights
- Live recognition → Sign prediction: predicts sequence
- Sign prediction → On-screen output: renders sign
- On-screen output → Operator: shows result