Taking a decent photo on your smartphone is easier than ever, but if you want to graduate to the next level with a DSLR or mirrorless camera, there’s a lot more to think about. Factors range from focus to shutter speed, aperture to ISO, color range and more, and knowing how to balance them all requires practice. Now, though, a gadget named Arsenal — currently raising funds on Kickstarter — claims it can do all the hard work for you, taking the perfect photograph every time.
think of it like a supercharged auto mode
Think of it like a supercharged version of the “auto” mode in any point-and-click camera. These auto functions adjust just one or two settings (like changing the exposure based on how bright the scene is) but Arsenal examines far more — looking at 18 separate factors in total. Most significantly, it also uses neural networks to recognize the scene you’re looking at and compare it to a database of professional photographs. It then takes the camera settings used to capture these images and applies them to your camera. You can tweak them to your liking, then take the shot.
Ryan Stout, the creator of Arsenal, told The Verge over email that he thinks his product fills a gap in the market. Stout says camera companies have underinvested in auto modes as they’re usually disdained by pro photographers (for good reason, he notes). This means, Stout says, they haven’t kept up with the advances possible using AI and new hardware.
“Arsenal is one of those products that shouldn't need to exist, but the camera companies have left some really low-hanging fruit,” he says. Stout thinks the Arsenal will appeal to beginner and intermediate photographers who want to make the leap from phone photography, but are finding it difficult to get good shots with their DSLR. “Some of that is composition, obviously, but I noticed a while ago that settings played a big role in that as well,” Stout says.
“the camera companies have left some really low-hanging fruit.”
To create the Arsenal, Stout trained a number of a convolutional neural networks (or CNNs) to analyze images. CNNs are also used by self-driving cars to scan the road, but instead of looking for things like pedestrians and road signs, the ones built in to the Arsenal look for features and scenery like waves, faces, and snow. Stout says he drew on a number of public datasets to train basic feature recognition for his neural nets, and then used high-quality photos taken from Flickr to teach the Arsenal what good camera settings looked like.
This information is combined with sensors built into the gadget, including a vibration-sensing accelerometer that can tell whether the camera is on a tripod or held in the hand. (For more information on the factors Arsenal examines, you can check out this blog post from Stout or watch this video demo using a prototype.) The gadget itself looks like a slim battery pack and clips into the hot shoe mount on top of a camera. It connects via Bluetooth to an iOS and Android app, which can also be used to create time-lapses and long-exposure shots.
Of course, when it comes to a gadget like this, the proof is very much in the pudding. We’ll have to see what type of shots the Arsenal can produce to find out if it’s worth the $150 price tag. But Stout certainly seems correct in suggesting that there’s an eager audience waiting for this technology. At the time of writing, his Kickstarter campaign has attracted more than $700,000 from backers — way over the $50,000 goal. It might not be long before mainstream camera companies also decide this sort of tech is worth a shot.