Expected Way to Return Image in Tensorflow Serving

[rohan mahajan]

In tensorflow serving, what is the best way to return an image. I found

many examples taking in images but none returning them. I would think 

it would be tf.encode_png or something like that but I having issues here with that

https://groups.google.com/a/tensorflow.org/g/discuss/c/yzN6Q_WJ9LE

[John Davis]

Rohan,

I am interested in your work, but I can't answer your question directly.  However I can point you to some code in the tensorflow docs which might help.

There is an example on image classification which uses the fashionNist dataset.  It uses code that pulls images to the filesystem as gzip files, then loads it to tensors. How it does decompression and read from files into tensors is hidden under the covers of the example.

https://www.tensorflow.org/tutorials/keras/classification#import_the_fashion_mnist_dataset

In a similar vein mxnet/gluon uses this same dataset.  Its similar in that you download the gz and then there is a transform you apply that converts to tensor.  I am curious how they both work, but I don't know.

Lastly, I am using each of these examples in my weekly workshop on gcp and aws.  I meet with foks via google hangouts and we work together on code.  My next meeting is this sunday at 10am EDT for gcp and 11am for AWS.   If you wish to join us this sunday and share what you have learned or simply work with me to figure this portion out I'm happy to collaborate.  FWIW, up to this point we have gotten the actual ML work out of the way.

Here is how to get to the weekly workshop https://rtp-gcp.org/about.html

--
You received this message because you are subscribed to the Google Groups "Discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to discuss+u...@tensorflow.org.
To view this discussion on the web visit https://groups.google.com/a/tensorflow.org/d/msgid/discuss/69cbf46d-7df8-4e00-a046-5972313dfe93n%40tensorflow.org.

--

John F. Davis
6 Kandes Court
Durham, NC 27713
919-888-8358
Public Profile https://www.linkedin.com/in/netskink

Personal https://skink.net

rtp-aws user group https://rtp-aws.org

GDG Cloud RTP/rtp-gcp user group https://www.rtp-gcp.org

独树一帜