Skip to main content

Artist got automated

Here is my first Tensorflow run to convert images to different styles of famous painters

I am using the code that is available at https://github.com/lengstrom/fast-style-transfer

It uses specific versions of below libraries
!apt install ffmpeg
!pip install tensorflow==0.12.1
!pip install numpy==1.11.2
!pip install scipy==0.18.1
!pip install pillow==3.4.2
I started with latest version and that triggered failure after failure.

If you are like me, add below changes to their respective files


#This is to make tensorflow behave as if it is in v1
File : evaluate.py and ./src/transform.py
import tensorflow.compat.v1 as tf
tf.disable_v2_behavior()




#Image functions are removed from scipy.misc module. so use imageio instead
import imageio
scipy.misc.imsave(out_path, img) --> imageio.imwrite(out_path, img)
img = scipy.misc.imread(src, mode='RGB') --> img = imageio.imread(src, pilmode='RGB')


Refer this post on more details on imageio changes

Here is the original image and after style changes











Comments

Popular posts from this blog

Netbeans 6.0 - Code generation

The Java editor is capable of generating often used constructs for you automatically. Press Alt+Insert to invoke the code generation menu and pick what you want to generate To add import statements: Use error hints : Click on the error mark with the light bulb or press Alt+Enter . It will offer you a list of possible classes to import. Pick the proper hint and you are done Use the import class dialog : Put the caret into the name of an unimported class and press Alt+Shift+I . A list of possible classes to import will appear. Use the smart fix import : Press Ctrl+Shift+I . You will get a dialog that lists all unresolved identifiers in the source. If there is more than one option for resolving the identifier you may choose using the combo box. Classes shown in gray do not fit for some reason. Method exit points : Putting the caret on the return type of a method definition will highlight all places where the method can exit Exception throwing points : Putting the caret on an...

My algorithmic Trading journey

Like many i used to manually watch and trade stocks, ETFs. Its non-productive and involves our emotions in trading decisions. So I wanted to automate trading. Initially used Trailing Stop and Limit orders. But they can automate only in a prescribed way. I started thinking about a way to trade programatically, where we have more control. My quest pointed to Zipline, a python algorithmic trading library. It support backtesting (testing algorithm with historic data) and most importantly supports live trading In this blog, I will walk you through my algorithmic trading journey. Zipline Installation From  https://www.zipline.io/index.html , I see it only supports python 2.7 or Python 3.5 Since python 2 was sun set on Jan 1 2020, there is no point in trying it. I went ahead with Python 3. For programming, installed latest Ubuntu 19.10 in a virtual machine . Ubuntu 19.10 comes bundled with Python 3.7.5. You can check your python version by typing python3  -V ...