Image Colorization Using GANs | Deep Learning | TensorFlow | Python

[@coderzden]

Please like 👍, Share 🗣 and Subscribe 🙌

[@chrisp.784]

Thank you so much! I am working on a thesis project related to picture colorization, your video saves me a lot of time!!

[@coderzden]

Thank you very much, it means a lot 😊

[@phanindardadwai8960]

Hello, are you done with the project?

[@baravkarprem6412]

This was really a superb implementation of GAN. How can we save this model into a h5 file? can you please make a video on it how to save and load GAN models?

[@coderzden]

Sure...trying my best to do it soon 🔜

[@peliculasperdidasyencontra5059]

Can you help me build a computer that will colorize?

[@thuongnguyenvan4743]

Thanks for the tutorial, I got a question though.. I want to save the model so I coded "generator.save". But i can't load the model. Tks

[@coderzden]

No worries 👍 Yes, you are doing it right way. I would need to have a look at your implementation to help you.

[@ttymmyc1121]

Really great work here. But I get one question, how to draw a curve on training/validation loss in this notebook? Thanks

[@coderzden]

Thank you very much, as per my understanding you can draw the curve by storing the losses in a particular data structure or you can use the checkpoints and then by using them you can make the curve. Thank you

[@ethos4066]

Thanks for the tutorial, I got a question though.. Where in the code do you train the discriminator twice? I see that you call it once just as you called the generator once.

[@coderzden]

Thank you very much, Yes you have figure out right, discriminator would be trained along side generator for same number of times. They both would be called once, one after another in the train_step() function, it's just their architecture which differs only.

[@YaroslavaMykhailenko]

Hello! Really good video, thank you! I have a question, can we output a metric - accuracy on each epoch when training a model? If yes, could you explain how to implement, I really interested)

[@coderzden]

Thank you 😇 Yes workout a bit with the training function to print the loss...

[@adithyareddy4382]

it is showing "Sorry, the file you have requested does not exist" Dataset is not available to download Please upload the dataset,update the new link

[@coderzden]

Dataset link has been uploaded in the description. Please like 👍, Share 🗣 and Subscribe 🙌

[@ycl1999]

How long does it take to train the model? Can colab run successfully without an upgrade?

[@coderzden]

Yes colab would be able to run it successfully...might take around 3 to 4 hours.

[@SameerJanu-gq6vn]

Really a good project and wonderful work ..but unable to open in github and collab notebook.can you provide new links

[@coderzden]

Aren't the file links in the description or comments working for you?

[@aureusyp]

@@coderzden the link of dataset, please update

[@ganapathisubramanyamjayam]

great work sir, could you please make a implementation video on super gans?

[@coderzden]

Thank you very much, it means a lot 😊 OK, I would try to make one.

[@selvamanisaravanan4773]

how do i measure accuracy

[@coderzden]

Using the loss/error given by discriminator (whether the given image was discriminated as actual or non actual by the discriminator)

[@anuran6180]

Hello Sir - Can you provide me the link for your dataset I am trying to open your Collab Notebook but it is opening with a white screen so it will be helpful if you provide me the link

[@mehmeteskandari3776]

the dataset link doesn't work

[@coderzden]

Sure, you may use the GitHub link to download the notebook locally : github.com/OvaizAli/Image-Colorization-using-GANs

[@coderzden]

@@mehmeteskandari3776 Sure, you may use the GitHub link to download the notebook locally : github.com/OvaizAli/Image-Colorization-using-GANs

[@anuran6180]

@@coderzden Yes sir I did the same but whenever I click on the download option of the dataset it redirects me to the Google drive showing file doesn't exist

[@coderzden]

@@anuran6180 Dataset link has been placed in the video description. Please like 👍, Share 🗣 and Subscribe 🙌

[@AnaghaKP-ou7su]

Please give the link of this dataset.

[@lalitjoshi1669]

The colab file is not opening, i want that file can you provide me uppdated link

[@coderzden]

It's available in the comments please check that out.

[@jagadeshwaranar3553]

Is this good enough to submit as a final year project.

[@coderzden]

Not really, you would be required to enhance it in several ways.

[@wallet_killer]

hey, how to save the model after training?

[@coderzden]

It's using the save() method of the Keras Model object

[@preethamk8010]

Can we upload our own black and white images and test it?

[@coderzden]

Yes you can, just update the data source file.

[@betsythomas5971]

Great work!!

[@coderzden]

Thank you very much, it means a lot 😊 Please like 👍, Share 🗣 and Subscribe 🙌

[@anya3662]

Hello! Thank you for sharing the codes and the training dataset. Excellent work! I got a question though. After training the model with the dataset of yours, how can I input my own black and white photo into the model to test the result? Thank you

[@coderzden]

Thanks for the appreciation. You would need to save the trained model and query it with your new input image. Search out for this process in TensorFlow. Thank you

[@amnaal5777]

Great and useful work, Is it possible to upload the dataset file and the github link please?

[@coderzden]

Thank you very much, it means a lot 😊 Dataset link has been uploaded in the description. Please like 👍, Share 🗣 and Subscribe 🙌

[@gauravghayal7941]

We gave our presentation fafter refering this video n sir told it completely wrong that unet isne used in it and generator image is wrong

[@coderzden]

Not sure specifically but the results and papers I referred all turn out to be as a good implementation and I was good to go with it. Please feel free to mention further queries here....

[@seyityahya5543]

Tez çalışmam için bakıyordum allah razı olsun başkan işime yaradın mk

[@coderzden]

Thank you 😇

[@thamielmoufid6816]

merci le code source svp

[@coderzden]

Welcome 😇

[@darsh558]

hi can i pleasee get a report on this project i need it for my college

[@HelloBud-j2q]

is this an pretrained model?

[@gowthamgowtham-bz2yr]

I have some doubts about the networks u built for the generator could u explain the way u built the networks for the generator please?

[@coderzden]

Yes sure you can clear your doubts. Generator network basically follows an encoder-decoder structure and it is being explained here 4:48. Moreover, if you want to deep dive how it's implemented in the code please have a look at this 7:33.

[@ritwiksinha7432]

Wonderful work. However, I have a question, the dataset which you have provided contains colour images, so from where does generator is getting Grayscale images? Thank you.

[@coderzden]

Thank you, actually I converted the coloured image dataset to grayscale while image processing.

[@ritwiksinha7432]

@@coderzden yes, but the main purpose of using GAN to colorise grayscale image doesn’t get resolved. Is there any project that you can guide where we have grayscale image from the beginning and as an output we get colorised image of it

[@abhiram-t5x]

Dataset link expired can you send the dataset please

[@محمدالجداوي-ج1ص]

have you got the dataset?

[@appanamukesh7264]

Google collab note book is not opening bro.. Kindly provide the updated link.

[@mubeenkodvavi6308]

Amazing project 😃

[@coderzden]

Thank you 😊

[@CrazyProgrammer16]

Good job friend

[@coderzden]

Thank you 😇

[@coderzden]

Dataset link has been placed in the video description. Please like 👍, Share 🗣 and Subscribe 🙌

[@AnikethVm]

hi can you please provide the data set link it has been removed. its really urgent as its for my college project

[@Varunmaniappan]

can we train these colorizing gan's using distributed computing framework like pyspark? if yes then how?

[@coderzden]

Yes they can be trained either through pyspark ML libraries or customizing on your own to the pyspark code

[@ame6184]

How do you use different gray image for the model to colorization instead of test_x ?

[@coderzden]

I split the dataset for training and testing at the start of the code