Creating Surreal Short Films From Machine Learning

Ever since we first saw the nightmarish artwork produced by Google DeepDream and the ridiculous faux paintings produced from neural style transfer, we’ve been aware of the ways machine learning can be applied to visual art. With commercially available trained models and automated pipelines for generating images from relatively small training sets, it’s now possible for developers without theoretical knowledge of machine learning to easily generate images, provided they have sufficient access to GPUs. Filmmaker [Kira Bursky] took this a step further, creating a surreal short film that features characters and textures produced from image sets.

She began with about 150 photos of her face, 200 photos of film locations, 4600 photos of past film productions, and 100 drawings as the main datasets.

via [Kira Bursky]
Using GAN models for nebulas, faces, and skyscrapers in RunwayML, she found the results from training her face set disintegrated, realistic, and painterly. Many of the images continue to evoke aspects of her original face with distortions, although whether that is the model identifying a feature common to skyscrapers and faces or our own bias towards facial recognition is up to the viewer.

On the other hand, the results of training the film set photos on models of faces and bedrooms produced abstract textures and “surreal and eerie faces like a fever dream”. Perhaps, unlike the familiar anchors of facial features, it’s the lack of recognizable characteristics in the transformed images that gives them such a surreal feel.

[Kira] certainly uses these results to her advantage, brainstorming a concept for a short film that revolves around her main character experiencing nightmares. Although her objective was to use her results to convey a series of emotionally striking scenes, the models she uses to produce these scenes are also quite interesting.

She started off by using the MiDaS model, created by a team of researchers from ETH Zurich and Intel, for generating monocular depth maps. The results associated levels inside of an image with their appropriate depth in relation to one another. She also used the MASK R-CNN for masking out the backgrounds in generated faces and combined her generated images in Photoshop to create the main character for her short film.

via [Vox]
In order to simulate the character walking, she used the Liquid Warping GAN, a framework for human motion imitation and appearance transfer, created by a team from ShanghaiTech University and Tencent AI Lab. This allowed her to take her original images and synthesize results from reference poses of herself going through the motions of walking by using a 3D body mesh recovery module. Later on, she applied similar techniques for motion tracking on her faces, running them through the First Order Motion Model to simulate different emotions. She went on to join her facial movements with her character using After Effects.

Bringing the results together, she animated a 3D camera blur using the depth map videos to create a less disorienting result by providing anchor points for the viewers and creating a displacement map to heighten the sense of depth and movement within the scenes. In After Effects, she also overlaid dust and film grain effects to give the final result a crisper look. The result is a surprisingly cinematic film entirely made of images and videos generated from machine learning models. With the help of the depth adjustments, it almost looks like something that you might see in a nightmare.

Check out the result below:

Continue reading “Creating Surreal Short Films From Machine Learning”

WiFi Goes Open

For most people, adding WiFi to a project means grabbing something like an ESP8266 or an ESP32. But if you are developing your own design on an FPGA, that means adding another package. If you are targeting Linux, the OpenWifi project has a good start at providing WiFi in Verilog. There are examples for many development boards and advice for porting to your own target on GitHub. You can also see one of the developers, [Xianjun Jiao], demonstrate the whole thing in the video below.

The demo uses a Xilinx Zynq, so the Linux backend runs on the Arm processor that is on the same chip as the FPGA doing the software-defined radio. We’ll warn you that this project is not for the faint of heart. If you want to understand the code, you’ll have to dig into a lot of WiFi trivia.

Continue reading “WiFi Goes Open”

Build Your Own Grid Tie Inverter

Inverters that convert DC into AC are pretty commonplace, some cars even have standard AC receptacles in them for you to plug in your favorite appliance. However, there’s a particular type of inverter called a grid tie inverter that allows you not only to make AC, but also inject it back through an AC outlet to power other devices in conjunction with the normal AC service. Why? Maybe you want to use your own generator or solar power. In some cases, the power company will pay you if you produce more power than you consume. Maybe you just want to know you can do it. That seems to be the motivation behind [fotherby’s] build, which is quite substantial.

The setup only handles about 60 watts, but it does all the functions you need: DC to AC conversion as well as phase and voltage matching. Actually, just converting DC to AC is almost trivial if you don’t care about the waveform. But in this case, you do care that you can create an AC signal to match the one already on the line.

Continue reading “Build Your Own Grid Tie Inverter”

Cheap Speakers Sound Good With Easy Open Baffle Design

If you’ve spent any time around audio gear at all, you’ll know that enclosure design is as critical as the speaker drivers themselves. [Frank Olson] demonstrates this ably, with his open baffle design for some cheap off-the-shelf speakers.

[Frank]’s aim was to do a comparison between using no enclosure, and an open baffle design, with a pair of 2″ full-range speakers. These drivers are nothing special; just a low-cost part that you’d find in any cheap set of computer speakers. [Frank] screws the drivers into a thin, flat wooden board, and then adds a supporting strut to allow the speakers to stand on their own.

The comparison makes it clear that even this basic baffle design makes a big difference to perceived sound quality. Bass is fuller, and the sound is far improved thanks to the baffle blocking out of phase sounds from the rear of the speaker.

It’s a technique that could prove useful to anyone quickly trying to rig up an audio setup for the workshop or makerspace out of leftover parts. We’ve featured similar projects before that espouse the benefit of enclosure design when using even very affordable speakers. Video after the break.

Continue reading “Cheap Speakers Sound Good With Easy Open Baffle Design”

Tarot Machine Flips Through Fate’s Rolodex

Were tarot card readers deemed non-essential in your part of the world (and do you think they saw it coming?) More than ever, we all need diversions that are for entertainment purposes only. And what better basis for entertainment than a mystical fortune-telling robot that can read your tarot cards?

This fantastic-looking ‘bot stands on the shoulders of [Scott Bezak]’s trailblazing method for easy DIY split-flap displays. Push the rather inviting-looking button on the top, and the flaps start flipping around to find your fortune. Once the fates have aligned, a thermal printer on the front spits out an image of your card along with an interpretation.

It’s obvious that [i_mozy] put quite a lot of effort into this slick machine, and we think the stickers look especially great. All the details of physical tarot card readings are accounted for, including a random number to decide the card’s position, and LEDs to represent the card’s element. Suspend your disbelief and check out the demo/promo video after the break.

Split-flap displays are a great choice no matter what you want to show. We’ve seen them used to display everything from the weather to the current Spotify track.

Continue reading “Tarot Machine Flips Through Fate’s Rolodex”

Books You Should Read: The Design Of Everyday Things

With everything from APIs to Raspberry Pis making it even easier for us to create and share objects shaped by personal whim, it’s high time that Don Norman’s sage design advice falls on not just the design student, but the hardware hacker and DIY enthusiast too. Grab yourself a coffee and a free weekend, and settle into the psychology of people-struggling-how-to-use-that-widget-they-just-purchased in The Design of Everyday Things: Revised and Expanded Edition.

Who’s to blame for a door that opens with a pull when everything about how it looks says it should open with a push? In Don Norman’s world, it’s not you; its the designer. Enter a world where blame is inverted and mistakes can be critically categorized. Norman takes us example by example showing us how common items in the world poorly serve the needs of their user, mainly because the designer simply ignores key aspects of our humanity. This book is a crisp, concise overview of human psychology when applied to engaging with things combined with a language of ideas to help us apply this psychology to better interactions. (And it reads like butter!)

Opening Up to the Language of Design

What’s an affordance, you might ask? Well, simply put, it’s a way that an object can be used by a human. How about a signifier? That’s a communication “signposting” scheme that object uses to suggest to you how it should be used. If that sounds a bit fluffy, just think about the last time you tried to push open a door that needed to be pulled. Something about that door was suggesting that you could push it open, but it couldn’t! It “fooled” you because all the object’s signifiers were telling you otherwise.

But Don Norman goes beyond a vocabulary that inverts our understanding of how we engage with objects and gives us another fresh perspective on how we make mistakes with out devices. Once again, these errors aren’t something to be ashamed of, but are categorizable interactions with our devices that, once understood, can be designed to accommodate or designed out altogether. Errors actually come in two large categories: mistakes and slipsMistakes are, by and large, errors in planning, and slips are errors of action. Have you ever set your alarm for 7PM when you meant AM? That’s a slip. Or perhaps you forget some items on your grocery trip? That’s a mistake But there are actually multiple subcategories, each clearly explained with examples from real life, often accompanied by disastrous consequences that may have been preventable with different design choices. Norman’s language for understanding mistakes is precise. And with this precision, we too can unpack everyday “mistakes” into a systematic way that lets us understand why they happened and how to mitigate or prevent them.

Here lies the power of the book. It’s a grammar book, one that teaches us the language of designers. Armed with the grammar of design, we can start to see the choices of designers and start making some thoughtful design choices ourselves.

A Refreshing New Look

Once you read this book, I’ll warn you. Though you may be armed with a new language, be careful with your criticism when you re-enter the world beyond that comfy armchair and empty coffee cup. Yes, in a way, this new vocabulary feels like a clever way to point a finger at “bad design.” And sure; with these new words and clearly articulated descriptions, we can do that. But let Don Norman do the blaming for you. This book is already riddled with examples of bad design drawing from either history or Norman’s personal experience. Instead, let’s put it to good use. The Design of Everyday Things is an opportunity for us as creators to reflect on how we communicate, how we suggest experiences, to the people who use our creations. So let’s make sure those experiences are good ones.

Side Note: the Revised and Expanded Edition of this book reads very differently from the original edition released way back in 1988. I strongly suggest finding the latest version if you can help it since so many of the examples have been brought up to speed with our times.

Hackaday Podcast 069: Calculator Controversy, Socketing SOIC, Metal On The Moon, And Basking In Bench Tools

Hackaday editors Mike Szczys and Elliot Williams march to the beat of the hardware hacking drum as they recount the greatest hacks to hit the ‘net this week. First up: Casio stepped in it with a spurious DMCA takedown notice. There’s a finite matrix of resistors that form a glorious clock now on display at CERN. Will a patio paver solve your 3D printer noise problems? And if you ever build with copper clad, you can’t miss this speedrun of priceless prototyping protips.

Direct download (~65 MB)

Places to follow Hackaday podcasts:

Continue reading “Hackaday Podcast 069: Calculator Controversy, Socketing SOIC, Metal On The Moon, And Basking In Bench Tools”