Your smartphone has a great camera, these apps will make the photos you take even smarter.
When it comes to taking a good photograph, the human eye is
still the best sensor. What happens to the image after that is probably
something a machine can do just as well or better, especially given the massive
number of images piling up in our photo albums today. And that's why the
breakthroughs in the next era of imagery are going to be done by computers.
Apps are already arriving that help us explore the possibilities, ranging from
retouching tools to software that addresses the enormous availability of images
in the digital age.
Here are five photo apps that are changing the way you think
about pictures.
Social Sweepster
We are taking more photos than at any other time in history.
Billions of images are being uploaded, many never to be looked at again. And
they might just include something that you don't want the rest of the world to
see. But who has time to scour a photo archive looking for the stray bong or
worse?
The idea that we need to focus on limiting—rather than
creating—images is one way in which our ideas about photos are changing in
2014.
A service that neatly explores this concept is Social
Sweepster. It's a tool that scans your online social presence (currently
Facebookand Twitter, but soon expanding to Tumblr and Instagram) and flags
questionable images that you may not want in the public domain.
“Our primary user would be someone who has recently
graduated from college and is looking to clean up their photos ready for job
applications,” says founder Tom McGrath. Recruiters regularly look at Facebook
and Instagram accounts as part of their employee screening process. One recent
study even demonstrated that it’s possible to predict job performance based on
the pictures on a person’s Facebook profile. That hilarious picture of you
passed out at your end-of-year college party? Maybe not so funny now.
As with many next-gen smart photo tools, Social Sweepster
doesn’t just look at the image itself to gather data. In addition to computer
vision, the software also uses text recognition algorithms to sift through
keywords associated with images. It’s even possible to examine the context of
images, since metadata can regularly reveal where a photo was taken.
“We’re really trying to tackle one of the hardest computer
vision problems out there, which is recognizing images in the wild,” McGrath
continues. “Recognizing a single beer can in a photo that’s small,
low-resolution, and badly lit is a real challenge. Being able to do that—and do
it accurately—is very, very tough.”
TouchRetouch
Removing a stray plastic bag from a photo of people praying
at the Ganges might get you disqualified from a National Geographic contest,
but for everyone else it simply makes for a better picture.
Whether it’s removing a photo-bombing stranger from that
lovely shot of you and your family, or taking out an ugly hotel from an
otherwise stunning landscape scene, one popular request for photo apps is the
ability to touch up existing photos. Unlike other methods of removing unwanted
detritus from pictures, TouchRetouch intelligently carries this work out on
your behalf, rather than requiring time-consuming manual work. Just select the
image component you wish to remove, and leave it to the software to do the
rest. Once an element has been selected, the app smartly analyzes what is going
to be required to fill a certain area, and then sets about filling it using
image components cloned from other parts of the photo.
The end result is impressive—and developer Kostyantyn
Svarychevskyy credits it with the new processing power of smart devices, which
can now carry out the kind of intensive graphical work that previously would
have required a much larger graphics-oriented machine.
“Increase of computational power of smartphones provides the
possibility of using new technology or advanced algorithms and user
experience,” he says. “In newer versions we [also plan to] try to improve this
technique on more complex backgrounds, such as buildings.”
Vhoto
The idea that the massive quantity of images we gather today
opens up new possibilities for photographers is the idea behindVhoto. “We talk
about camera ubiquity a lot as a team,” says creator Noah Heller. “What does it
mean when everyone carries devices with multiple cameras built into them? And
what happens when those cameras are on all the time? You have to ask yourself
what you’re going to do with this amazing amount of content.”
Vhoto uses computer vision technology to scan your videos to
find and extract the best photographic moments. “The concept that you have to
press a button to take a single picture is a really old idea that goes back to
chemical cameras,” Heller continues. “That no longer has to be the case. If you
want a record of a great moment in your life, why not just let the camera go
and then let technology sift out and sort the best end images. Our mantra is
that users should think of photography as fishing with a net, not with a hook.”
Some of the metrics Vhoto examines are fairly
straightforward: sharpness, clarity, color, and the presence or absence of a
face or smile. But the model also takes into account more abstract features
like novelty, context, and composition. Rather than the photographer having to
be consciously aware of all these elements, the app learns preferences based on
the past behavior of individual users so it gets better at predicting what
photographic elements you’re likely to be interested in. Whatever pictures you
end up sharing, saving, or otherwise interacting with will be analyzed so that
future similar images can be elevated within the model.
“It’s not our job to force people to like photos a
professional critic might say is better composed, it’s our job to help people
get the photos that they want,” Heller says. “If our users turn out to like
photos with a certain color composition or facial expression, that’s what our
machine learning model needs to deliver.”
Color Thief
The rise of Instagram has made filters increasingly popular,
but some tools take the concept of post-processing pictures further than
others. Color Thief is an example of a great color correction app that should
be on every budding smartphone photographer’s device.
“Color Thief takes the colors from one of your photos and
transfers them to another,” says creator Aaron Barsky. Blurring the line
between functional image modification and something entirely new, Barsky likens
the app to challenging a painter to repaint your photo, using only the color
palette from another photo of your choosing.
“We count how often a color is used in both photos,” Barsky
continues, explaining how the app functions. “We then transfer the most
frequently used color from the source to the most frequently used color in the
target, and similarly down to the least frequently used color.” The challenge,
he says, is in grouping colors together. “A photo could have hundreds of subtle
shades that a human would identify as light blue—but the computer sees as
completely different colors. We use ‘mathemagic’ to make sure the color
transfer happens with smooth gradients of color.”
Although Color Thief is a post-processing step for images,
rather than a camera app in itself, it still benefits from the improved quality
of smartphone cameras. “Color Thief works best on photos that have a sharp
in-focus foreground with a blurred background,” Barsky says. “As our users can
take better and better photos with the built-in camera, the more fun they'll
have remixing those photos with our tool.”
AutoStitch
It started as a computer vision research project at the University
of British Columbia, and now AutoStitch is a panoramic photo app that leaves it
rivals in the dust. It has two major benefits over other similar apps, as well
as the built-in panoramic functionality found in an increasing range of
smartphones.
It's a versatile tool that doesn't require taking a single
sweep shot. As long as the images overlap in some way, the photographer is free
to experiment with images in any order or arrangement—including horizontal,
vertical, or a mixture of both. It’s even possible to stitch together photos
taken with different camera apps, as well as those imported from other devices.
The quality of the finished images is also vastly superior
to other panorama apps. Inputs are composed of full resolution images, which
allows for each photo to be composed individually. The overlapping regions of
these high-def photos are then automatically blended to ensure seamless
transitions between images. The end result is an impressively professional
panoramic photograph.
“By using the other sensors on board, and with the sheer
processing power available, the door is open to create tools that will take
smartphone cameras beyond what is possible with traditional cameras in many
ways,” says developer Geoff Clark, speaking about the future of smart camera
apps in general. “Augmented reality shooting guides that analyze the images in
real time, or light-field capture that allows for re-focus of images, are a
couple of examples.”
So go ahead and snap all the photos you like. Just put them
somewhere accessible to the algorithm that's going to make them worth looking
at again.
Source:
Posts
