There is not a lot to do in my cave, so I got satellite internet and have been spending a lot of time messing around online. Hello, Steemit! I also love to draw. Recently I have been combining these two pastimes. I have discovered that doodling a small (sticky note sized) drawing in cross hatch style can yield some amazing results when scanned in at high resolution and ran through the Deep Dream Generator. The neural network can apply detail to make the drawing look appropriately detailed at much larger sizes than the original doodle. Last night I drew a quick sticky note portrait of my cousin Horny Cyclops.
Original pen and ink Horny Cyclops on sticky note:
| Below is an example of raw output from the Deep Dream Generator, using the image to the right as style image. Note I do not own rights to this style image, I am presenting here as fair use in reporting on my process. |
|
Sometimes I like the resulting image as is, but more often I take the image output from the Deep Dream Generator and remove the background in PhotoShop to just leave the subject. In the case of the post cover image, I also set the transparency to 75% so the original pen crosshatch would show through a little bit.
Here is an example of one with the Deep Dream Generated background:
PhotoShop Inner Bevel effect used here to make this seem like a gold medallion:
Another example of setting the image output from the Deep Dream Generator to 75% transparency in PhotoShop and overlaying the original pen and ink cross hatch so that some of the original pen lines show through:
