This is a challenge is Very interesting, I use the Deep Learning and Torch deep learning framework to accomplish my task with and without using GPU.
What it does
This is painting two more images to create unique merging using deep learning with torch. The first image is the original. The later images are the generated results, with the painting from which the style was sampled shown in miniature.
How I built it
Very First Step I Create VM in OpenPOWER Super Vessel Cloud with Ubuntu Virtual Machine in the cloud. Next creating the instance in Spark with all the cuda, spark, hive, hadoop ready Virtual machine. Using VPN configuration I used to configure the VPN for my local access using putty ssh client. for CISCO mobility VPN control will help to done the steps. once i connected in Cisco VPN login. Then i logged into my VM and followed some sequence of steps to makes my VM as ready for deep learning project(you-net) categories. Configured my own deep learning project, which the Neural deep learning painting. For that, first I installed the Torch 7 in my VM then Lua and Luarocks and caffe via Luarocks.
you will find a deatiled Once i Successfully configure my VM the i RUN my project as simple as following with and with out GPU.
with GPU Enabled $ th neural_style.lua -style_image starry_night.jpg -content_image tubingen.jpg -output_image tubingen_starry.png -gup -1
without GPU Enabled $ th neural_style.lua -style_image starry_night.jpg -content_image tubingen.jpg -output_image tubingen_starry.png -gup -0
Challenges I ran into
The challenges in this to configure the machine with all required packages. the torch makes lots of time to do installation in My VM running in OpenPOWER cloud.
Accomplishments that I'm proud of
Finally i successfully installed and configured my VM Machine with all the need software like Torch 7, Lua, Luarocks, Luaijit, caffe images, NN etc. some installation will makes some panic but i do it.
What I learned
The Open POWER cloud comes with variety ranges of VM images and some of then are already ready with some package like CUDA and lot more. The Spark service especially BIG data analysis service, database service are incredibly with more powerful CPU cores, highest RAM memory and very high performance Disk.
What's next for Neural-Painting-Torch
The next version is to make this more faster with all other deep learning frameworks.