How cristian-bejarano/contando-dedos works
Lengua de Señas Colombiana is a webcam-based sign-letter recognition prototype. The documented path is to capture labeled hand-motion samples, train a classifier, and recognize signs live or in a guided practice mode; evaluation scripts compare model performance. The excerpts show capture and recognition using MediaPipe landmarks, normalization, and persisted Random Forest artifacts. The README mentions a menu entry point that is absent from the tree, so it is not mapped; evaluation-script relationships are shown only where source explicitly supports them.
Main entry points: reconocer_libre.py Live recognition · reconocer_practica.py Practice recognition · capturar.py Capture samples · datos Labeled landmarks
Components 13
Sample collection
- Capture samples
abcedario/capturar.py - Labeled landmarks
abcedario/datos
Model lifecycle
- Train classifier
abcedario/entrenar.py - Model artifacts
abcedario/moldes - Evaluate classifier
abcedario/evaluar.py - Repeated CV analysis
abcedario/inferencia.py - Evaluation reports
abcedario/evaluaciones
Recognition experience
- Live recognition
abcedario/reconocer_libre.py - Practice recognition
abcedario/reconocer_practica.py
Vision runtime
- Hand landmarker
modelos/hand_landmarker.task - Landmark demo
prueba_mediapipe.py
Other components
- Signer
- Webcam
Connections 21
- Signer → Capture samples: performs signs
README.md - Webcam → Capture samples: supplies frames
abcedario/capturar.py - Capture samples → Labeled landmarks: writes samples
abcedario/capturar.py - Labeled landmarks → Train classifier: loads samples
abcedario/entrenar.py - Train classifier → Model artifacts: saves model
abcedario/entrenar.py - Model artifacts → Live recognition: loads model
abcedario/reconocer_libre.py - Model artifacts → Practice recognition: loads model
abcedario/reconocer_practica.py - Webcam → Live recognition: supplies frames
abcedario/reconocer_libre.py - Webcam → Practice recognition: supplies frames
abcedario/reconocer_practica.py - Live recognition → Signer: shows letter
abcedario/reconocer_libre.py - Practice recognition → Signer: shows letter
abcedario/reconocer_practica.py - Evaluate classifier → Labeled landmarks: loads samples
abcedario/evaluar.py - Evaluate classifier → Evaluation reports: writes reports
abcedario/evaluar.py - Repeated CV analysis → Labeled landmarks: loads samples
abcedario/inferencia.py - Repeated CV analysis → Evaluation reports: writes distributions
abcedario/inferencia.py - Capture samples → Hand landmarker: configures
abcedario/capturar.py - Live recognition → Hand landmarker: configures
abcedario/reconocer_libre.py - Practice recognition → Hand landmarker: configures
abcedario/reconocer_practica.py - Landmark demo → Hand landmarker: configures
prueba_mediapipe.py - Signer → Live recognition: uses mode
README.md - Signer → Practice recognition: uses mode
README.md