Signapse Helen Keller

🔥 🔥 First Production Signapse Release! 🔥 🔥

Check out Signapse v3.0 Helen Keller! Our first production-ready release adds a bunch of features:

• Our setup.sh script has been tweaked to massively reduce setup time on all platforms! #10
• Our GUI experience has been tweaked to allow a threshold parameter to be set per-sign. Users may adjust the threshold to suit learning needs and the network latency on their target platform. #25
• The code-base has been meticulously refactored with a focus on thread safety, memory safety and making objects adhere to SOLID design principles #22 #24. Our code has matured to be production-ready and easy to modify for future extensions.
• Documentation has been added via DocStrings inline with the code, this is accessible via our online hosted Docs, and built with our continuous GitHub action #11
• Further unit tests have been added into our existing test library. Our unit test coverage is wider than ever and is continuously called with our GitHub action. #4
• Our Wiki is now populated with all sorts of useful information ranging from installation guides to deep-dives into the Signapse video processing architecture and neural network selection. #23
• We have extended our reach over social media channels, shot a demo video, and even gone viral on TikTok! #12 #27

Signapse v3.0 is awesome! We are so excited to share our software and would love to hear about your experiences learning to sign! 🥳

Helen Keller

Signapse Thomas Gallaudet

Check out Signapse v2.0-beta!

This release adds a bunch of features:

• Our video processing architecture has been re-designed to make use of callback routines (#15). This allows for much better object encapsulation and a more modular code-base.
• A User interface has been wireframed (#3) and integrated into the codebase (#16 #18). Built in Qt, the GUI allows users a much more intuitive experience interacting with Signapse and shows live feedback in real-time as well as a progress-bar to make learning more fun.
• 29 unit tests have been written using the google-test framework (#4) these are called with our continuous integration github action to ensure code integrity throughout development (#13).
• We have added our python scripts used for training the neural network used (#9). This allows transparency with network generation and translation to a format suitable for the OpenCV runtime. We hope that making our training scripts public will allow for members of the community with CNN expertise to help develop the next generation of sign-language models.
• Known bugs in the software have been addressed (#14).

There's lots more to come, please check out our new release and let us know how you get on!

Thomas Gallaudet

Signapse Pedro Ponce de León

The first release of Signapse. This is a proof of concept and celebrates the achievements of the team thus far. Following the milestone set out for this release, the software includes:

• Automated integration testing
• Beginnings of a vibrant wiki and documentation for the code base
• Trained CNN architecture for recognition of hand signs in real-time
• Software components for acquiring video feed from the camera and processing with our neural network
• A basic user experience which sets challenge signs and gives the opportunity to practice each sign with GUI feedback

Signapse still has many features to be added, and its fair share of current bugs, but this release has all the basic components to get our ideas our there in the open. Please try out Signapse and let us know how you get on!

Pedro Ponce de León