Hand Sign Language Recognition System -- 2
Budget: ₹1,500 – ₹12,500 INR
I need a software that captures hand signs through a mac integrated webcam, identifies each alphabet in real time, and streams the detected characters on screen so they build words one letter at a time (e.g., showing “H I T H E R E” as the user signs to make it display as HI THERE with each alphabet while detecting should also be heard as audio).
Core expectations
• Webcam video feed is analysed frame-by-frame.
• Every recognised alphabet appears instantly on screen, forming complete words without manual intervention one by one with a time interval to make next sign.
• An optional audio cue should pronounce each detected letter so users hear what they just signed. A basic text-to-speech library is fine.
Technical preferences
I am comfortable with either C++ or Python—choose whichever delivers faster, more reliable results. The underlying tech stack is flexible: OpenCV for image capture, plus TensorFlow or PyTorch if a lightweight model is needed. Feel free to reuse publicly available datasets or pretrained models as long as results remain accurate and latency stays low on a mid-range laptop.
Deliverable
A runnable script or small application with clear instructions, commented code, and a short read-me explaining how to run the software on macbook. I’m aiming for a functional software with good accuracy and efficiency keep the scope tight and focused on core sign-to-text (and optional voice) functionality.
Core expectations
• Webcam video feed is analysed frame-by-frame.
• Every recognised alphabet appears instantly on screen, forming complete words without manual intervention one by one with a time interval to make next sign.
• An optional audio cue should pronounce each detected letter so users hear what they just signed. A basic text-to-speech library is fine.
Technical preferences
I am comfortable with either C++ or Python—choose whichever delivers faster, more reliable results. The underlying tech stack is flexible: OpenCV for image capture, plus TensorFlow or PyTorch if a lightweight model is needed. Feel free to reuse publicly available datasets or pretrained models as long as results remain accurate and latency stays low on a mid-range laptop.
Deliverable
A runnable script or small application with clear instructions, commented code, and a short read-me explaining how to run the software on macbook. I’m aiming for a functional software with good accuracy and efficiency keep the scope tight and focused on core sign-to-text (and optional voice) functionality.