The original Fawkes ( https://github.com/Shawn-Shan/fawkes ) project came on my radar when I have started working on adversarial examples in deep learning. Fawkes cloaks facial images by adding pixel-based perturbations to the original images, barely visible to the human eye. This bypasses face-recognition systems (or at least weakens their confidence).
I highly recommend reading the original paper available from the developer's website: https://sandlab.cs.uchicago.edu/fawkes/ to understand how it works. Generally speaking, Fawkes computes _cloaks_ by maximizing feature similarity to unrelated faces while minimizing DSSIM (Structural Dissimilarity).
These cloaks are then applied to the original images, producing cloaked images. These cloaked images look as similar as possible to the original images, whereas their feature space deviate "as maximal as possible" from the original.
Having an opinion on something doesn't make the speaker a self righteous person.
Regarding the topic, I assume that with "the hords of spaghetti monsters" you mean lack of software perfection. In our profession, it's mostly about how to manage the lack of software perfection, and the answer does not always have to be a microservive. They are great for controlling hardware resources in a cloud-deployment, but then I see tech-dudes dockerizing everything anytime without asking why.
It's a decision every dev team have to make for themselves.
Opinions are like asses - everyone has one. Self righteous people are those who think that only their opinions matter and try to present them as facts, rolling over any reasonable objection with hand waving and insulting the other party for not understanding things.
Thank you! I didn't even knew that Show HN existed. It's exactly the right thing.
I'm currently comparing the latencies for recording (recording through a web browser vs recording through an android app) and seeing how much I can reduce the latencies. Rockr wants to be useful, and useful in audio engineering generally means "low-latency" (especially in a loopstation where audiotracks must align perfectly). So it's not sure at that point for me to either stay in the web browser or switch to a potential Rockr-app.
JQuery offers optimizations, but they are limited to the capabilities of web browsers of course.
I am building a music loop station for the web browser. Right now it's just an audioplayer with some audio effects, but the goal is to make it a handy tool for music-practitioners and (maybe in a far future) a loop station for live performances.
Regarding 1: I remember the internet of the 90s and find our current internet comparatively boring in terms of UX. Back then, websites were created manually, without using "abstractions" like wordpress or wix. This forced us to think freely about UX and to hack HTML+CSS as much as possible. The browser's interface you mentioned is quite powerful. And with HTML6 on the way it becomes even more powerful.
Regarding 2: A paramount program utilizing responsiveness at it's finest won't take compromises for different screen sizes, but it takes longer to develop. Compromises like these usually emerge through the economic circumstances of software development.
2. You will quickly end with two programs with different features, or maybe a single program that only offers a fraction of features depending on input/output capabilities :
Screen size is just one aspect of this, consider how different programs might be optimized for a vertical screen layout, or the use of smartphone's accelerometers, or working best with precise mouse movement, or relying on the user learning to use keyboard shortcuts (ideally with an easy on-ramp phase) for effectiveness, or relying a lot on hovering your mouse over features for popups (which is very poorly translated as a press and hold on touchscreens - consider how rare hover text for pictures has become).
(More extreme examples would be also being able to use the software from, for instance, a monochrome, low refresh rate watch with only a few buttons, at which point you get the option to use vibration feedback notifications that you can't expect to have on a desktop.)
I highly recommend reading the original paper available from the developer's website: https://sandlab.cs.uchicago.edu/fawkes/ to understand how it works. Generally speaking, Fawkes computes _cloaks_ by maximizing feature similarity to unrelated faces while minimizing DSSIM (Structural Dissimilarity).
These cloaks are then applied to the original images, producing cloaked images. These cloaked images look as similar as possible to the original images, whereas their feature space deviate "as maximal as possible" from the original.
DISCLAIMER: I am not the developer of Fawkes. I merely developed a web-interface for it. Big thanks to the [original researchers from SANDLAB, Chicago: https://people.cs.uchicago.edu/%7Eravenben/publications/abst...
Yes, the trained model is part of the repo. Just in case you're irritated by it's size (~300MB).