photoriot

Alignment makes random pics work together more-predictably.

I want my pics to generally have any vertical line in the center of the frame be aligned vertical. What I've been doing is manually rotating and cropping, but I've got too much backlog to catch up on. At the same time, I'm sold on the concept of mirrorless now, and it seems like a mirrorless cam should be able to rotate and crop live in the viewfinder and when saving the photo.Failing that, I'd want the tilt recorded in metadata so I can automate the rotate/crop on the computer. Does the RP or other R's record tilt? Ideally, I'd like to see a maximal crop rotated in the viewfinder. I'm told the 5D's record tilt, so fallback is to get a beater to take a bunch of reference pics to train neural nets to detect the angle.

http://phobrain.com/pr/home/gallery/pair_horiz_blinds_shadows_ofc_bench_armrest.jpg

Street photography.

Done! :-)

The url got hidden, demos.algorithmia.com - probably it could be much improved with effort, and I assume there is much better proprietary stuff used for colorizing moves.

Have you power-cycled everything?

Google has wrapped their deep learning image recognizer into a package you can play with fairly easily if you can handle console access to a computer ("TensorFlow for Poets"). You simply create a folder for each thing you want to recognize, and fill it with examples (~600 pics per flower in the flower classification example they provide), and run a command for 5 mins to a day (2012 laptop). Then you run another command to get the classification for a new image. The deep parts of the net remain fixed, and were trained by Google on a big database of pics - it just fine tunes the final stage to your pics. Here are my first results - pretty shaky for my own purposes, but not bad for a start. I have different requirements than Google, e.g. if I say the photo has red or white, I mean it jumps out at you, not simply that it is there, so to get the most out of this technology I might need to get a GPU or 4 and retrain the entire network for a week or two, and manually classify way more than the 15K images I have now as input to the algorithm.

Note how it gets Weston's cabbage red. Colorize Black and White Photos http://edward-weston.com/wordpress/wp-content/uploads/2015/01/Cabbage-Leaf-1931-39V.jpg https://a.1stdibscdn.com/archivesE/upload/a_115/1485891672037/johnny_edward_weston_1944_weston_gallery_carmel_california_l.jpg http://edward-weston.com/wordpress/wp-content/uploads/2015/01/Pepper-1930-30P.jpg http://edward-weston.com/wordpress/wp-content/uploads/2015/01/Nahui-Olin-1923-16PO.jpg

I haven't read the article in full, but suspect the author could benefit from reading this: Photo Editing with Generative Adversarial Networks (Part 1) | Parallel Forall

Another concept that the photos don't support, again planned meticulously long in advance of just loading the pics in my manual pair formatter.

Can you believe it took me 3 days to add a 'Let pics repeat' option to Phobrain? This demo pair seems like something of a match for the pics above.