TensorFlow Tutorial 2: Image Classification Walk-through

TensorFlow Tutorial 2: Image Classification Walk-through

GitHub repo:

[ethangreenhouse]

This is the clearest and simplest tensorflow image classifier tutorial I could find. I scoured all of the internet and mostly found bs videos like "Make a image classifier in 5 mins" and got stuck and confused by the second step. Thanks for making a real tutorial!

[MRDJR]

Thank you, this is the best tutorial on Youtube by far. Much easier to understand than even Tensorflow's own tutorial.

[siddhantindave]

Thank you very much, my friend! I was trying to create a classifier for a long time and nothing except your method helped me acquire 99% accuracy! You are the best!

[elinak]

Thank you very much. After modification of the code and running on tf 1 all works!!! Big Like

[binaypradhan]

Hello Chris, Thank you for sharing the tutorial. With the latest version of TensorFlow the sample project does not seem to work. If there is any new updated version of the repo then please share in your comments.

[jacobusstrydom]

The best example ever!!. I know the theoretical part but always struggle to get stuff to work. Thanks!!

[mdshahjalal]

Best Tensorflow tutorial. A thousand of wishes of the new year 2019

[poornarao.vattikuti]

finally one working model ..! thanks dude

[michaelmutekeri]

dude that awesome..everything works fine and we can see it work..looking forward for the next video

[aymericperonnau]

Hello
I want to use mobilenet instead of inception for the training to deploy the test on a raspberry pi.
I changed the architecture to mobilenet_1.0_224 in the retrain.py code. The retrain code works put then when I run the test, u have this message: Cannot interpret feed_dict key as Tensor: The name 'DecodeJpeg:0' refers to a Tensor which does not exist. The operation, 'DecodeJpeg', does not exist in the graph.
I don't know how to fix it and I have to finish the project before Thursday. Does anyone know how to fix it ? Thank you

[yacine]

Thank you . Please i need help, Why test prgm test only for one image it does not test the other images in test_images ..

[aberamiranjan]

Hi Chris, I got an error when I try to run retrain.py " ImportError: DLL load failed: The specified module could not be found ".Nothing seems to solve this issue.Please help..

[cobyreimer]

Thank you for making these tutorials they are pretty in depth. However I followed the first, and the second tutorial video. And I get error messages whenever I try to run the retrain.py file, it fails to import numpy as np, i tried changing and deleting the line, turns out it is not correclty importing anything. I have no clue how to fix this does anyone have any ideas?

[trieulieuf]

Hi Chris, i downloaded and run your repo. It identifies road bikes and mountain bikes correctly with great confident. But i throw a random girl picture to test_images folder, and it identify her as a bike as well, with great confident too (a little lower than the actual bikes).
So how can i classify her as "not a bike" ?

[feyzadogan]

Hi Chris, there is no retained_labels.txt file so program gave an error
starting program . . .
ERROR: RETRAINED_LABELS_TXT_FILE_LOC does not seem to exist

[adrianmanuelorayeabrasaldo]

Hi Chris, is it possible to train a fire/flame image? just like any other objects that are being trained to be detected.

[koetschkid]

Hi Chris: Great video, this got me started on the right track with a working image classifier. I am working now with my own training images trying to identify digits within coin dates. I have run into an issue with classifying 6's vs 9's as they are almost identical just simply rotated 180 degrees. All the other digits work fine. Is there a way to ensure that the classifier recognizes the images only in a certain orientation? All of the training images used are in the proper orientation already. I noticed that there are some methods within TensorFlow such as but I think using those might make the situation worse. Any suggestions on a way to remedy this issue would be greatly appreciated.

[arnaudberard]

Hello i’m trying to figure out where this problem comes from, after running and implementing more than 500 photos i get on pycharm : ERROR: there are less than the required 10 images in can someone help ?

[JoanneHastie]

Thank you so much for tutorial 1 & 2. This has gotten me the farthest into getting tensorflow working on my computer. I was originally using Atom as my editor and running in the command prompt window in my windows OS. When I noticed test.py did not keep going after the first image (I cannot tell in the tutorial with bikes if you are closing the image to get the next result image to pop up?) I have tried both closing and leaving it alone to see if the next test image with result will appear but nothing happens. So I installed Pycharm, thinking it was an issue for working in the command prompt. But the same thing happens. How do you get to review the next image? Thanks again!

[andrewhogan]

Thank you for this! Everything is laid out in such a thought-through manner. One followup question to this: If I wanted to export the results of the Image classifier test.py to a file (csv, txt, excel or any other) any idea how I would go about that? I tried using logging but that didn't yield any useful results .