Recognize if an image contains an object from a database

40 views
Skip to first unread message

Giali BoM

unread,
Aug 3, 2020, 11:01:02 AM8/3/20
to TensorFlow.js Discussion

I'm working on an online merchandise recognizer I would like to upload a photo and make my web page recognize if the object is one of those present in my database, or if the object being uploaded is my product.

I have seen that the thing can be solved with Tensorflow.js, but I'm really out of this world and I need some advice to find my way.

is this possible?

In my own small way I opted for tensorflow.js precisely because the page is written in Javascript / Html / CSS like all web pages

I tried, for object detection, to consider the Teachtable Machine https://teachablemachine.withgoogle.com/ I have seen that it is possible to export the model for tensorflow.js, but I don't know how to add the result on my page!  

Thank you

Jason Mayes

unread,
Aug 3, 2020, 2:03:12 PM8/3/20
to Giali BoM, TensorFlow.js Discussion
Your best option is to use Cloud Auto ML which will make it easy for you to upload all the images of your products neatly in different folders to Google Cloud Storage. You probably want a few hundred or maybe even thousands of each object you want to recognize (and be sure to include different types of the same thing eg colours etc), taken from different types of camera, in different lighting conditions, and in different environments to achieve best results.

Cloud AutoML can then have a chance to use this data to produce a machine learning model to classify future images. It may take several hours to train but once ready you will be given option to export to TensorFlow.js format which will then allow you to use the resulting model on your website directly if you wish. 

This is probably the most robust production method right now for you short of writing your own system in Node.js using TensorFlow.js itself which would require much more understanding of machine learning to do.

Jason

--
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.
To view this discussion on the web visit https://groups.google.com/a/tensorflow.org/d/msgid/tfjs/8ef3864c-3fc2-4829-bc87-95e9b2b3d81en%40tensorflow.org.


--




     
  •  Jason Mayes
  •  Senior Developer Advocate, TensorFlow.js
  •  Google
     Follow my discoveries or contribute your own

   

This email may be confidential and privileged. If you received this communication by mistake, please don't forward it to anyone else, please erase all copies and attachments, and please let me know that it has gone to the wrong person.


Fu Chung-Yi

unread,
Aug 4, 2020, 1:03:52 PM8/4/20
to TensorFlow.js Discussion

Giali BoM

unread,
Aug 5, 2020, 2:46:05 AM8/5/20
to TensorFlow.js Discussion, Fu Chung-Yi
Thank you soooooo much!!
This is perfect!!!
I will never stop thanking you 

Giali BoM

unread,
Aug 5, 2020, 2:46:56 AM8/5/20
to TensorFlow.js Discussion, Jason Mayes, TensorFlow.js Discussion, Giali BoM
thank you, in case I need a more precise and perfect solution than the other one they suggested, I will follow these instructions
Thank you
Reply all
Reply to author
Forward
0 new messages