tinrab / go-tensorflow-image-recognition

Image Recognition API in Go using TensorFlow
MIT License
250 stars 77 forks source link

How did you get the sources #11

Open ZZ0R0 opened 1 month ago

ZZ0R0 commented 1 month ago

Hello,

Very nice work,

How did you do that ?

Where did you find the source code ?

Thank you ;)

tinrab commented 1 month ago

What do you mean exactly?

ZZ0R0 commented 1 month ago

Hello,

First of all I'm glad you replied me,

I didn't expected a reply of yours

My question was a bit messy because I was tired.

To clarify it my goal is to use MoveNet model using tensoflow in golang or more recently this repo => https://github.com/wamuir/graft/releases?page=1

Actually my goal is so to convert an image from gocv to bytes and then to a tensor

I tried importing your code to convert an image to tensor but it produces a float tensor and Movenet expects a in32 tensor from the image's Bytes

What i was asking is how did you created the function to convert the image to a tensor by using tensorflow "op" which i guess stands for operations , especially, the makeTransformImageGraph function ...

If i'm rigth this was buit for tensorflow version 1.X but the movenet model is only available in 2.X

To be more clear, how did you built the makeTransformImageGraph function ?

Did you reverse engineered the convert_to_tensor function from the official tensorflow source code ?

Btw thanks for reading my messages and again, nice work on github keep it running :)

Best regards,

Robin

Le mer. 31 juil. 2024, 14:52, Tin Rabzelj @.***> a écrit :

What do you mean exactly?

— Reply to this email directly, view it on GitHub https://github.com/tinrab/go-tensorflow-image-recognition/issues/11#issuecomment-2260452656, or unsubscribe https://github.com/notifications/unsubscribe-auth/AUFUTDCAIKFYVJTFLLE4XFDZPDMXRAVCNFSM6AAAAABLXF4EFCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDENRQGQ2TENRVGY . You are receiving this because you authored the thread.Message ID: @.***>

tinrab commented 1 month ago

This repository is 6 years old, and I think I looked through TensorFlow's source code for a solution. Sorry, I can't you with that.