SignTrack is a sign language transcriber. It analyzes, processes, and recognizes sign language in real-time, with exceptional accuracy and efficiency. From transforming signs to text to helping you learn sign language through its application in interactive learning, SignTrack helps make computers more open for everyone by taking a human-centric approach to computing.
SignTrack is a sign language transcriber. It analyzes, processes, and recognizes sign language in real-time, with exceptional accuracy and efficiency. SignTrack helps make computers more open for everyone by taking a human-centric approach to computing.
<br/>
...
...
@@ -29,7 +29,7 @@ SignTrack utilizes a state-of-the-art LSTM model that predicts based on a sequen
## Showcase
## Data Collection
### Data Collection
Making the data collecting experience friendlier for the user has been one of our top priorities. That's why we made sure to create an easy-to-use data collection user interface, even for those with minimal coding skills.
...
...
@@ -37,7 +37,7 @@ Data plays a fundamental role in creating a great model that's both accurate and
Another feature of SignTrack Data Collect is flipping the image to generate data as if you signed with the other hand too! Creating a model that can make equally accurate predictions on both hands.
Privacy is at the center of SignTrack. All of the collected data is free of personal data, like raw images or computer specifications. Making users feel more comfortable exchanging datasets.
Privacy is at the center of SignTrack. The collected data is free of personal data, like raw images. It only stores numerical values of the keypoint positions as NumPy arrays, making users feel more comfortable exchanging datasets.
<palign="center">
<imgsrc="Assets/readme/Collecting1.gif">
...
...
@@ -73,7 +73,7 @@ The consistent, unique, and identifiable design continues on the main program wh
## Dataset
SignTrack comes with an expansive dataset, consisted of 193,945 collected keypoint sets for ASL. The given model, which is trained on this dataset, is recommended to be used on web cameras on or above head height. If the given model does not perform as expected you can always add more data using DataCollect. When creating your dataset remember to make sure to try signing from different angles and positions to create a more generalized model, that will work on a broader spectrum of angles.
SignTrack comes with a dataset, consisted of 193,945 collected keypoint sets for ASL, and can be easily extended. If the given model does not perform as expected you can always add more data using DataCollect. When creating your dataset remember to make sure to try signing from different angles and positions to create a more generalized model, that will work on a broader spectrum of angles.
