jahnvijain777/traffic-sign-recognition-project ? reverse-engineered prompt

Build me a web app called SignSight that can पहचान traffic signs from an image upload and show the most likely result with confidence scores.

I want it to work right in the browser, so the prediction happens locally with no backend needed. The user should be able to upload a traffic sign photo, see the image preview, and get the top prediction plus a few alternative matches out of the 43 supported traffic sign classes. Make the interface simple, modern, and responsive, with a clear results area and loading state while the model is running.

Please wire it up so it can load the trained model from the public model folder, preprocess the image the way the model expects, and handle errors nicely if the model file is missing. If you need to check current TensorFlow.js or Vite docs online, go ahead.