Surreal-ly into a Deep Dream

This was an accidental selfie. I was in my favourite spot for the afternoon light beside TheWindow. I stopped to finish my coffee, right in the shadow of the door frame, and didn't realize I set off the self-timer on my camera from my phone. The actual photo was lame but I decided to try it out in Google Deep Dream Generator, using a "modern art" style pattern.

I think the results are somewhere avant garde 1920ish.

PhotoProject :: My #HandDrawnPhoto approach to #AIart

The success of the Belemy portrait auction, has encouraged me to compare Obvious’s DAG approach to my use of Google Deep Dream algorithm. Deep dream uses a convolutional (foreword chaining) neural network to find and/or enhance patterns in an image. This can be applied in two ways Style & Inception (google’s terms). The inception process tries to find the characteristics of a given object/pattern within a photo and tends to produce surreal images. Style on the other hand  seeks to put characteristic patterns back into a photo based on features the trained neural network has recognized. This is the technique I have become very interested in as a way to put my characteristic marks (and sometimes colours) back into a photo. Hence the name Hand Drawn Photos for my approach. The method does have superficial similarities to Onvious’s DAG Approach.

Here I am using my sketch for inktober as the image to be processed and a photo of the same trees as my training image used to build the neural network (AI). The Photo here is acting like the discriminator in Obvious’s approach. The result (below) is tonally similar to the photo and kind of a messy squiggly version of something I might sketch.  But…

In this, my preferred approach, I am using the photo as the image to be modified and my sketch as the image to train the neural network and letting the system modify the photo to “find” my characteristic mark making (aka style) and modify the photo. In this case I am taking the colours from the photograph, but that is not essential.

I feel this second result is much cleaner and fresher, closer to something creative and artistic, Although I am strongly guiding the outcome by using my own sketches (as opposed to an automated algorithm). So the big question is “Do you consider this #HandDrawnPhoto to be legitimate AI art? I’m warming to answering yes.

If you are interested here is the link to my profile and public images on Google Deep Dream website. (Its free to join and use)

Love to hear your views,  please leave your comments here.

PhotoProject :: Building a “Hand Drawn Photo” in 3 layers

I felt it was time to do something creative for this blog (and instagram). I had selected this photo of the young Osprey on the nest on the Mooloolaba spit to make a painting or perhaps a pastel drawing. However I felt it also need to be reinterpreted as a “hand draw photo” using a combination of rendering approaches.

First I loaded the original photo into Google Deep Dream Generator and used the Abstract Nature trained AI filter. This created a darker rendering but I loved the strong blue colouring of the sky. Next I took this new version into Corel Painter as the work to base a clone painting using the artist impressionist clone brush and a linen surface texture and also emphasised warm colours or the plumage and nest.

For the Final Work (above) I used On1 Photo RAW and load the original photo as the base layer and the Clone painting as the second layer. Then I used soft light blending mode and used a semi-opacity masking, a lot like a homemade luminosity mask to bring out some of the original photo (eg almost 100% masking over the eye and beak, but variable 75% or lower on selected parts of the nest and feathers). I also lifted the vibrancy of the blue and yellow/oranges a fraction to give me a strong complimentary colour feel. Finally I cropped in a little to add a bit more drama to the composition.

Back to experiments in HandDrawnPhotography

I’ve been sketching at the beach and also took a photo (or two) so it is a good time get back into the Hand drawn photo project. which is an elaborate way to say I’m merging my sketching with a digital photo. In this examples I’m using google deep dream generator  (in Deep Style Mode) and using my hand drawn sketch a the training image. I’m using the mark making style from the sketch but trying to maintain the colour from my photo.

 

Deeper into Disappointment with Google

Over the years google has bought onto the internet a number of great software tools. I have and still use many of them, Blogger, picasa, gmail, My Tracks, Google Earth, Maps, Google Drive (Google Docs, Slides & Sheets), Nik Software, Snapseed). I’ve gone right off google photos (but still put up with it to publish photos into this blog). A number of these have been depricated and whilst they still work google no longer offers downloads, or alternative software and/or no support. The one thing they have been good with to date is giving warning and letting you download any personal data.

I have taken I liking to the idea of exploring that place between art and a photograph, in particular what I have been Hashtagging #AIart (on Instagram). Google created the Deep Dream generator originally as a way to explain how their image analysis technique worked. It is based on training neural networks (a type of artificially intelligent data miners). As well as recognizing objects, animals, people & things the networks can also be used to seek out style (ok in a rudimentary way, it can mimic marks making, line works, colour palette and several other features that make artworks original. The software does not create original art but it can take a photo and find that style of linework & colour etc within a photos. Some of the result are amazing. Unfortunately more and more examples that I see produced by other  “copy cat” system are degenerating into the garish and boring, as they become very popular of social media.

Clearly google are looking to muscle in of the social media aspect and they have add a multilevel aspect to site membership (its free) and that is tied to number of public post and more importantly number of likes. A different levels you can only generate a specific number of a specific type/style of image with a set time frame.  So It only took me a couple of images to reach my limit with a polite come back and try again later. The type of thing that just plain kills exploration and creativity. The writing is on the wall google this is going to become a ghost town like google+

The one nice thing, is the UI (user interface has been smarted up a bit with better layout of selection buttons and other tools. Nothing new just a little bit cleaner to look at.

One new feature, that I have asked for a few times, is better documentation of what was set up  to produce each result. It seems quiet common that most people (especially those new to deep dream) blanket bomb one image with many alternative networks/styles, and after doing several and waiting a decent time for each it is difficult to remember how you created a particular masterpiece. The ? Intro item now gives you a run down although it has only remembered a few of my style images.

Perhaps its another case of needing more likes?

No matter, such is life.

So Why am I complaining, the software tools are free?

Tell him he’s dreaming

This was one of those good ideas at the time that took much more effort than I anticipated. I’m going to blame google for changing its deep dream geneator, It just took a long time to do something I concieved in a flash. That was to redo my week of photos as abstract deep dreams

#AIart Shoot Out

I was certainly an early follower of the developments in the deep dream style art filters, which use neural networks to recognize and amplify aspects of any photographic images. I must admit I like the potential of the forward seeking neural networks that can replicate line work and other mark making characteristics of the training image. I  chose to this called Deep Dream Style, and starting giving such creations the  hashtag #AIart (which seems to have caught on on Instagram). My prediction that these might become over used Art Filters is probably also an astute observation, there is certainly a plethora of such apps. Fortunately most are based around famous artist and great art, Even Adobe is joining the crowd offering researching a AI style filter that can “copy” the look from another photo. I still think these filters can be used creatively rather than as “look at me” plagiarism so obvious in social media today.

This photo was just an experiment, the first photo I took with my new HTC U Play mobile phone. I was trying out the AutoHDR feature and it handled the difficult strong lights and shadows well. Its just afternoon tea, aka Coffee & fresh figs. It is an interesting enough composition so I thought I would try the image across a variety of current #AIart tools (google deep dream style, on computer, Dremscopeapp & Prisma on phone)

Which rendering do you like best?