technology:image compression

  • Women Once Ruled Computers. When Did the Valley Become Brotopia? - Bloomberg
    https://www.bloomberg.com/news/features/2018-02-01/women-once-ruled-computers-when-did-the-valley-become-brotopia

    Lena Söderberg started out as just another Playboy centerfold. The 21-year-old Swedish model left her native Stockholm for Chicago because, as she would later say, she’d been swept up in “America fever.” In November 1972, Playboy returned her enthusiasm by featuring her under the name Lenna Sjööblom, in its signature spread. If Söderberg had followed the path of her predecessors, her image would have been briefly famous before gathering dust under the beds of teenage boys. But that particular photo of Lena would not fade into obscurity. Instead, her face would become as famous and recognizable as Mona Lisa’s—at least to everyone studying computer science.

    In engineering circles, some refer to Lena as “the first lady of the internet.” Others see her as the industry’s original sin, the first step in Silicon Valley’s exclusion of women. Both views stem from an event that took place in 1973 at a University of Southern California computer lab, where a team of researchers was trying to turn physical photographs into digital bits. Their work would serve as a precursor to the JPEG, a widely used compression standard that allows large image files to be efficiently transferred between devices. The USC team needed to test their algorithms on suitable photos, and their search for the ideal test photo led them to Lena.
    0718P_FEATURE_BROTOPIA_01
    Lena

    According to William Pratt, the lab’s co-founder, the group chose Lena’s portrait from a copy of Playboy that a student had brought into the lab. Pratt, now 80, tells me he saw nothing out of the ordinary about having a soft porn magazine in a university computer lab in 1973. “I said, ‘There are some pretty nice-looking pictures in there,’ ” he says. “And the grad students picked the one that was in the centerfold.” Lena’s spread, which featured the model wearing boots, a boa, a feathered hat, and nothing else, was attractive from a technical perspective because the photo included, according to Pratt, “lots of high-frequency detail that is difficult to code.”

    Over the course of several years, Pratt’s team amassed a library of digital images; not all of them, of course, were from Playboy. The data set also included photos of a brightly colored mandrill, a rainbow of bell peppers, and several photos, all titled “Girl,” of fully clothed women. But the Lena photo was the one that researchers most frequently used. Over the next 45 years, her face and bare shoulder would serve as a benchmark for image-processing quality for the teams working on Apple Inc.’s iPhone camera, Google Images, and pretty much every other tech product having anything to do with photos. To this day, some engineers joke that if you want your image compression algorithm to make the grade, it had better perform well on Lena.

    “We didn’t even think about those things at all when we were doing this,” Pratt says. “It was not sexist.” After all, he continues, no one could have been offended because there were no women in the classroom at the time. And thus began a half-century’s worth of buck-passing in which powerful men in the tech industry defended or ignored the exclusion of women on the grounds that they were already excluded .

    Based on data they had gathered from the same sample of mostly male programmers, Cannon and Perry decided that happy software engineers shared one striking characteristic: They “don’t like people.” In their final report they concluded that programmers “dislike activities involving close personal interaction; they are generally more interested in things than in people.” There’s little evidence to suggest that antisocial people are more adept at math or computers. Unfortunately, there’s a wealth of evidence to suggest that if you set out to hire antisocial nerds, you’ll wind up hiring a lot more men than women.

    Cannon and Perry’s work, as well as other personality tests that seem, in retrospect, designed to favor men over women, were used in large companies for decades, helping to create the pop culture trope of the male nerd and ensuring that computers wound up in the boys’ side of the toy aisle. They influenced not just the way companies hired programmers but also who was allowed to become a programmer in the first place.

    In 1984, Apple released its iconic Super Bowl commercial showing a heroic young woman taking a sledgehammer to a depressing and dystopian world. It was a grand statement of resistance and freedom. Her image is accompanied by a voice-over intoning, “And you’ll see why 1984 won’t be like 1984.” The creation of this mythical female heroine also coincided with an exodus of women from technology. In a sense, Apple’s vision was right: The technology industry would never be like 1984 again. That year was the high point for women earning degrees in computer science, which peaked at 37 percent. As the number of overall computer science degrees picked back up during the dot-com boom, far more men than women filled those coveted seats. The percentage of women in the field would dramatically decline for the next two and a half decades.

    Despite having hired and empowered some of the most accomplished women in the industry, Google hasn’t turned out to be all that different from its peers when it comes to measures of equality—which is to say, it’s not very good at all. In July 2017 the search engine disclosed that women accounted for just 31 percent of employees, 25 percent of leadership roles, and 20 percent of technical roles. That makes Google depressingly average among tech companies.

    Even so, exactly zero of the 13 Alphabet company heads are women. To top it off, representatives from several coding education and pipeline feeder groups have told me that Google’s efforts to improve diversity appear to be more about seeking good publicity than enacting change. One noted that Facebook has been successfully poaching Google’s female engineers because of an “increasingly chauvinistic environment.”

    Last year, the personality tests that helped push women out of the technology industry in the first place were given a sort of reboot by a young Google engineer named James Damore. In a memo that was first distributed among Google employees and later leaked to the press, Damore claimed that Google’s tepid diversity efforts were in fact an overreach. He argued that “biological” reasons, rather than bias, had caused men to be more likely to be hired and promoted at Google than women.

    #Féminisme #Informatique #Histoire_numérique

  • Super Tiny Website Logos in SVG
    https://shkspr.mobi/blog/2017/11/super-tiny-website-logos-in-svg
    https://github.com/edent/SuperTinyIcons

    You may not realise it, but #bandwidth is expensive. It costs you time, money, and battery power whenever you download a file larger than it needs to be.

    That’s why I’ve become a little bit obsessed with #SVG - Scalable Vector Graphics. They’re the closest thing to magic that the web has when it comes to image compression. Let me show you what I mean.

    This is the standard Twitter #Logo. It’s 512 * 512 pixels and, even with hefty #PNG compression, weighs in at around 20KB.

    Here’s the same logo rendered as an SVG. Because it is a vector graphic it can be magnified infinitely without any loss of fidelity.

    The uncompressed SVG is a mere 397 Bytes. Not a #typo. You could fit over 3,000 of these images on a floppy disk.

    That’s why I have released SuperTinyIcons on GitHub. Eighty of the web’s most popular logos - each image is under 1KB.

  • Google launches #Guetzli, a new open source JPEG encoder that creates high quality JPEG images with file sizes 35% smaller than currently available methods

    https://research.googleblog.com/2017/03/announcing-guetzli-new-open-source-jpeg.html

    #compression #image_compression

    Guetzli [guɛtsli] — cookie in Swiss German — is a JPEG encoder for digital images and web graphics that can enable faster online experiences by producing smaller JPEG files while still maintaining compatibility with existing browsers, image processing applications and the JPEG standard. From the practical viewpoint this is very similar to our Zopfli algorithm, which produces smaller PNG and gzip files without needing to introduce a new format, and different than the techniques used in RNN-based image compression, RAISR, and WebP, which all need client changes for compression gains at internet scale.