I have 50+ tabs related to this topic and the worst writer’s block in years. So instead of writing a full-blown review of the current community efforts to generate images, I’ll give a short overview of my thoughts. To get a basic idea of what I am talking about, look at https://softologyblog.wordpress.com/2021/11/25/text-to-image-summary-part-5/ https://twitter.com/RiversHaveWings In the last 9 months, there has been a lot of enthusiasm in the NN community about image generation, specifically, text guided image generation....
Colorizing with old and new
I have recently got a chance to colorize a dozen photos of my grandfather. I wanted to document thre results here as I think they indicate an important pattern in practical AI. TLDR For personal use, choose complete apps and MyHeritage in particular. For industrial one, choose solutions optimized for your problem rather than the SOTA models and DeOldify in particular. When using the newest model, be prepared to spend months ironing out the kinks....
Using neurons and spotting patterns
“You’d be surprised how much gray matter is dedicated to the analysis of facial imagery. Shame to waste it on anything as … counterintuitive as residual plots or contingency tables.” Watts, Peter. Blindsight (Firefall) (p. 337). Tom Doherty Associates. Kindle Edition. Before reading any further try to find a pattern in the video. Is eye size related to the hair color or maybe the older the face the wider the nose?...
Closing the tabs
A beginning There is a lot of stuff on the Internet. When I find something interesting I tend to leave a tab open, as an unsolved task, as something to get back to. Way too often I never get back but the tabs remain open. Silent, judging, unread. And even when do read them often I need to experiment, to save results in order to fully process the info. Thus this blog....