v3.0 SiLaTra Improved Performance, Models, Code

• 12 Gestures support using HMM added
• 33 hand poses recognition using grid-based feature extraction followed by KNN classification improved
• Many minor bugs fixed
• Corresponding Android Code in Client Repo improved by huge extent
• Heavy reorganization of code and modularization
• Old deprecated codes archived
• Documentation added

Silatra Android App with Realtime Recognition of Digits and Letters

• Uses Tejas's Grid Model for classification
• Android app sends out frames to Python server which call c++ file for segmentation and then in Python the rest of the logic for classification is present.
• The server then sends back the modal value of a set of 15 consecutive recognized signs.

1st Prototype Presented

This is the 1st working prototype as demonstrated to the Mentor and Subject Matter Expert.
Features included:

• Image segmentation, feature extraction and classification using KNN / KMeans / SVM / Deep Learning / Random Forest

• RTree test implementation submodule implemented

• Skin color detection by Machine Learning implemented

• Interface for exchanging image between C++, Python implemented