Deep Photo Styletransfer copies a photo style into another

We already know we can't believe everything we see, and this Deep Photo Styletransfer is just one more reason for that, as it can create amazing realistic images that simply aren't real - though are based on real ones.

Google's AI can enhance 8x8 pixel images

A few months ago we saw Google's super-resolution upscaling in action, but not it's taking thing even further, adding detail and making sense of even low resolution 8x8 pixel images.

What a Tesla sees when it drives around

Tesla's new auto-pilot mode allows them to drive automatically with no human intervention. It does seem like magic but, as you might imagine, there are lots of technology working behind the scenes. Technology that you'll be able to glimpse in the video I bring you today.

Google can translate images into words

A picture may be worth a thousand words, but computers have a very hard time doing something even a 2 or 3 year old child can easily do: recognizing what's what in a photo. But we're getting there...

Wolfram Alpha shows us how a dog sees Obama

Wolfram Alpha is an amazing knowledge engine that can answer questions we can't even imagine. Today I bring you one such case. Would you imaging Wolfram Alpha to be capable of showing us how a dog sees Obama? Well... it may sound like an April's foll prank, but it actually can.

You can check it out for yourself, by asking Wolfram Alpha: apply dog vision to image of obama.

This system assumes the result based on two color photo-receptors on dog's eyes, and 20% of human visual acuity, and it gives us an estimate of how our trusted companions see us.


Next I tried doing the same for cats, but it seems Wolfram is more of a dog-person, as the same query for cats didn't provide any results. You can try other images besides Obama though, I'll leave that for you to explore.

More stuff you can do with Wolfram Alpha? How about...

• What's on the other side of the world
• What's the lifespan of different animals
• How many bills you'd need to cover a country

Amazing stuff... isn't it?

Visolu Similar Image Sorting for iPhone and iPad

Ever scrolled throught hundreds or thousands of photos on your iPhone or iPad, trying to find that particular photo you know you took on a white snowy mountain, or that deep blue sky, or yellow sandy beach? With this Visolu App for iPhone and iPad that's something that becomes a lot easier.

Instead of sorting your photos chronologically, this apps analyzes each and every one of your photos, and sorts them according to their visual similarity. That means that you can quickly zoom in the "redish" spectrum of your photos (maybe sunsets?), or the "greenish" part - probably plants, or whatever you like.

You can even search for images similar to the one you're currently looking at, not to mention that it can also show you where the photos were taken (on a map) if you had the GPS option enabled.

The best part is: this App is free for a limited time - so... hurry up and get it while you can.

Eulerian Video Magnification Reveals Unseen World

Sure, we all know just how amazing high-speed camera footage looks like, allowing us to make time slow down to a crawl and watch every tiny detail of thing happening in a fraction of a second.

But... what if there were a lot more going on, invisible to our naked eyes, but that could be brought to light using some clever techniques? That's precisely what happens with this Eulerian Video Magnification process.

It's an algorithm that can amplify not only spatial movement, but also things on the color and frequency domains.

Meaning, using this method, one can actually see the effect of the flow of blood in a human body.

As well as showing us our veins contracting and expanding as blood is pumped from the heart.

An amazing process, that makes us wonder just how much is out there, that is kept invisible... right in front of our eyes.

HDR Thermal Imaging

High Dynamic Range photography is now increasingly popular, thanks to smartphones and compact cameras. This is a technique that captures several shots with different exposures, and then combines a final image where over/under-exposed areas can be still have all the detail, as they're stitched together from the most appropriate frame.

This happens because image sensor aren't capable of capturing nearly as high a dynamic range (extremely dim/bright areas in a single shot) as our eyes can. And the same thing happens even when you're dealing with non-visible spectrum, like thermal cameras.

When you use thermal cameras you have to select wich target temperarure range you want to capture, meaning anything below or above it will not be distinguishable. Unless...

If you apply the same multiple-exposure HDR concept  to thermal cameras - or, as they call it "super-framing" - you can extend the thermal sensor temperature range considerably and ensure you have a high-resolution image with a wider thermal range.

You can think of it as quickly capturing a frame of 0-50Cº and another from 50-100Cº (and so on) in quick succession, and then assembling a final image where you can see all the fine thermal details.

Inserting CGI Objects into Photos

Though computer generated images can now be created with such realism that we can't really say if they're real or not, there's still a lot to be done when you want to insert a CGI object into a real-life photograph. Though you can put a lot of effort into it, it isn't easy to make something artificial look like it belongs there...  until now.

Check these amazing examples that show not only how you can insert digital objects into still pictures... but even create realistic animations that interact with its surroundings.


Rendering Synthetic Objects into Legacy Photographs from Kevin Karsch on Vimeo.

Realtime 3D Face Substitution

It's still a work in progress, but it serves to show us that in the not so distant future, one might better think twice about who we think we're talking to... even over a video-call.

As you can see in the following video realtime face substitution is indeed well within our grasp.


Faces from arturo castro on Vimeo.

Tracker Free Video Analysis

Ever wished you could track and analyze motion video, similar to what you see on TV when they "dissect" sports with digital overlays? Well, thanks to this amazing and free tool called Tracker you can do that and much more. Not only you can analyze all sorts of motion systems captured on video (like falling objects) and compare it to theoretical models, you can use it for much more than that.

If you have a camera pointed to any street, you can easily figure out how fast each vehicle is moving, or even figure out how fast you can throw/kick a ball. The possibilities are endless.



You can easily find tutorials and examples on YouTube (like this one) but the best part is, you can play around with it and figure it out by yourself. After all, it's a free video analysis tool.
(I only wished I had this sort of stuff when I had physics classes! :)

Javascript Face Detection

Here's another perfect example of what you can achieve by simply mixing together you can freely find around the web.

Daniel Goodwin took Mozilla's Rainbow, which gives your browser access to your computer webcam, and linked it to a javascript face recognition library.

The result is as follows:


Demo: Mozilla Rainbow + Canvas + JS Face Detection from Daniel Goodwin (@dsg) on Vimeo.

So, you still think it will take long for a website to recognize your face and greet you appropriately? :)

Portrait Professional - Easy Photo Retouching

I can't really say I'm a big fan of significantly altered photos (what's the point of retouching a photo to the point of "perfection", when real people are nothing like that?)

However, if done correctly, I won't argue that it can embellish some photos. If you're no Photoshop master but have the need to tweak portraits, you might want to give this Portrait Professional a try.

Basically, it's a simplified photo editing software, that allows you to quickly select the eyes, nose, mouth, face contours, etc. and try out different corrections and effects.


Portrait Professional retouching software - Gavtrain review from Gavin Hoey on Vimeo.

There are some "extreme" examples in their gallery, and I wouldn't suggest you push it that far, but... it does look like a nice tool for even an amateur photographer to master quickly.

Deblur your Photos

No matter how good image stabilization techniques are, it seems we cannot get rid of blurred photos!

But, thanks to some clever guys at Microsoft Research, things are about to get in focus.

Thanks to a motion tracking apparatus (6 DOF) that captures the movements made during the image aquisition thet can apply a corrective digital filter that can un-blur your photos.

Now, they only need to combine this with those "super-resolution" algorithms and I think it's safe to assume that will be the end of those blurry UFO/aliens/ghosts photos! :)

Photoshop CS5 Puppet Warp

It may not be as amazing as the magical fill I showed you the other day, but just like the real content aware fill in  Photoshop CS5, this puppet warp is sure to make digital image manipulation a lot easier to a lot of people.

3D Book flipping Scanner

How about a book scanner capable of digitizing a book as you flip through it?


That's what prof. Masatoshi Ishikawa at Tokyo University has done in his lab, using a high speed camera, strobe lighting and a laser iluminator in order to detect the 3D shape of each page as it turns, so it can then be reconstructed in a "flat perspective."

Image Reconstruction

Looks like those incredible image enhancements we see on TV are becoming increasingly real.
What if someone told you it was possible to reconstruct an image from a noisy or incomplete source?
No matter how impossible it sounded, someone applied some advanced math to it and... the results were amazing.

Be filling the empty spaces using the known pixel as reference, they were able to get crystal clear images out of incomplete MRI scans. And the same applies to most other kinds of images and even sounds!

And this is not all: how about compressing an image by deliberately throwing away the needless pixels that can be rebuilt using this algorithm? As you can imagine, the possibilities are endless!

Sikuli: Visual Image Search and Automation

Just found this amazing paper: Sikuli: Using GUI Screenshots for Search and Automation
From MIT's Tom Yeh, Tsung-Hsiang Chang, and Robert C. Miller.

The idea is simple: how often you wished you had the possibility to use images instead of words?
They say an image is worth a thousand words, and sometimes that is actually the case.

In the graphical computer world we live in, there are things that just can't be translated into words, and using images would greatly simplify some tasks.

From their abstract:

We present Sikuli, a visual approach to search and automation of graphical user interfaces using screenshots. Sikuli allows users to take a screenshot of a GUI element (such as a toolbar button, icon, or dialog box) and query a help system using the screenshot instead of name.

But that's not all... imagine you want to create a script to automate some task on your computer, and you have to press an icon that may be positioned in different places?

Something like... this:

Sikuli also provides a visual scripting API for automating GUI interactions, using screenshot patterns to direct mouse and keyboard events. We report a web-based user study showing that searching by screenshot is easy to learn and faster to specify than keywords.

This kind of tools would allow you to easily do some complex image processing tasks, such as - for instance - a baby monitoring script, that would check if the baby was sleeping face up.

A really interesting project we should all keep an eye on: Sikuli!

PhotoSketch

Are you one of those people who can't draw or do anything "right" in image processing software?

Worry not, this PhotoSketch will save the day!

At last we'll have computers actually doing artistic work for us.
All you need is to draw a rough outline of what you want in your "perfect picture" and the computer will search for the best images and compose it all together. Although there are still errors that need ironing out, most of the time the results are amazing.


PhotoSketch: Internet Image Montage from tao chen on Vimeo.

PatchMatch and Video Seamcarving

Do you remember that amazing resizing method called seamcarving?

Looks like Photoshop CS5 will take that even further by using a new patchmatch plugin that makes complex element moving a trivial task. Just see for yourself:


PatchMatch: Structural Image Editing from Dan Goldman on Vimeo.

And if you think these thing only work in static images, just check this video seamcarving examples to see how we'll soon be able to rescale video at will: