Tensorflow JS and tiny yolo

[Srinivas Ramakrishna]

We are using tfjs yolo tiny for image detection. THe offiical yolo model link is https://modeldepot.github.io/tfjs-yolo-tiny-demo/

We have integrated this with an web application where on page load we are downloading the model and later the user scan the image via the webcam and if its a valid image then they will go to next page.

The issue we are facing is mainly on the IOS devices. The model trained is around 10mb and when we scan the object on any IOS device the browser will get freezed and becomes unresponsive.

Is there any optimization that can done either on TFJS or on the model to increase the performance and run smoothly on the ios device?

Also are there any other model that is light weight which can be used with tfjs and works smoothly on iphone?

[Nikhil Thorat]

I think the TensorFlow.js version is quite old on that demo, if you upgrade I bet it will be much more stable on iOS. Can you do that and let me know if you still have issues?

--
You received this message because you are subscribed to the Google Groups "TensorFlow.js Discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tfjs+uns...@tensorflow.org.
Visit this group at https://groups.google.com/a/tensorflow.org/group/tfjs/.
To view this discussion on the web visit https://groups.google.com/a/tensorflow.org/d/msgid/tfjs/fdc1eae7-1c6f-42da-a310-84bf7cc45874%40tensorflow.org.