Bigger output and server setup?

[Convolutionair]

I am looking into the fork from 'mtyka', as it has a 'hack' or way to deliver bigger output.

• Can I use a g2.2xlarge Amazon aws for example,
  to render big images: let's say 100 x 200 cm. on 300 dpi?
• If this is possible, how much time would it take for a render like this on a g2.2xlarge?
  (For calculating the avarage server costs.)

(Most big examples I've seen so far had around the 4000 pix with 96 dpi. So I am thrilled to see bigger outputs.)

Great work on this coding, keep it up!
Thanks in advance

[andersbll]

Hi, with the vanilla neural artistic style method I wouldn't expect you to get good results on images of that size. The span of the highest level convolutional features are around 32 pixels. Therefore, the content/style features you merge would be miniscule in an image with a width of 200 cm at 300 dpi (>20k pixels).

[Convolutionair]

Thanks for the info Anders.