Is Apple’s “Deep Fusion” image processing a Game Changer? Let’s see what happens when you put the iPhone 11 Pro with the new “Deep Fusion” beta up against a similarly priced DSLR.
I know what you are thinking. Why do we need yet another iPhone vs. DSLR comparison, where the DSLR features pro glass and has a full frame sensor? We don’t, and I’m not going to subject you to such a pointless comparison. What I will do is compare the new iPhone to something that is much closer in price, and far more likely to be equipment you might actually own, a Nikon D7200 DSLR. The Nikon features a 24MP crop sensor which is widely regarded as having excellent dynamic range (see DXOmark for details). I’ve been shooting with the D7200 for several years now and absolutely love it when I’m in the mood for the whole DSLR experience. It’s a solid camera that can be had for under $1000 new, often with a kit lens, and much less used. Sure, I could have pull out a Nikon D850 but that’s really not a fair comparison as many posts and videos keep demonstrating. At the end of the day I wanted to show how a $1200 iPhone with Deep Fusion did against a $1200 DSLR, not how it performs against something costing more than twice as much.
The Collings Foundation put on a great show and are genuine in their efforts to educate the public on the history behind some of history’s greatest moments.
In order to give each camera system a chance to shine I decided to take them out to a local event that featured both indoor and outdoor subjects. The perfect opportunity presented itself with the Collings Foundation’s “Battle for the Airfield” event in Stow MA. Being a World War II reenactment, I knew there would be plenty of action and static subjects including tanks, soldiers, and encampments to explore. The Collings Foundation also has a new tank themed museum which features some seriously rare WW2 tanks which I was excited to check out for the first time. If you have not yet had the opportunity to either attend one of their lawn events or even just a short visit to the museum, I highly recommend you add it to your bucket list. They put on a great show and are genuine in their efforts to educate the public on the history behind some of history’s greatest moments.
First, let me get something out of the way. Deep Fusion. What is it, and why does it change the game? For those of you who don’t want to spend an hour discussing the use of complex convolutional neural networks for detail enhancement I will just cut to the chase. Deep Fusion is one of the first working examples of what I would call “next generation computational photography”. Unlike pure image stacking, which has been used by photographers to eliminate noise and increase detail for years, Deep Fusion significantly ups the game by using machine learning algorithms to process both short and long exposure photos into a single high-detail final image. The true magic in the process is that the machine learning algorithm Apple uses has been trained to understand what the final result “should” look like and is instrumental in ensuring that your photos meet this expectation as much as possible. This is the primary reason you don’t see Deep Fusion being employed on photos taken in ideal lighting conditions. As the iPhone 11 Pro’s camera already products excellent photos when the light is good the computational requirement to further enhance these photos would basically be wasted. So Deep Fusion only engages when the lighting is less-than-ideal, and where you are going to see the most benefit from the technology.
So, what does this mean for our comparison? It means we might get a fair fight after all. It also means we should consider that any iPhone equipped with Deep Fusion may punch above its weight when it comes to image quality under less-than-ideal lighting conditions, such as what you might encounter inside of a museum (spoiler alert).
Let’s start with the outdoor photos, where I fully expected the Nikon to dominate, especially in the action shots where having the ability to slap on an 80-200mm f2.8 lens would give it a major advantage. I was not disappointed in this regard. The Nikon easily produced top quality images of the action, stopping motion thanks to a high shutter speed, and just mopping the floor with the iPhone.
The iPhone, lacking anywhere near as much optical zoom range, struggled to produce acceptable photos for anything beyond about 4x on the default camera app’s zoom dial. This was not unexpected and gives a lot of credence to those who say there is no replacement for displacement quality glass that has the right focal length for the job. Once the action stopped, and the distance to the subjects came on in, things began to change. The iPhone started to not only get competitive with the Nikon but thanks to Smart HDR (another “computational photography” feature), it even pulled ahead on some shots displaying richer colors and better overall out-of-camera image quality.
Once the party moved into the museum it was not even a contest anymore between the Nikon and the iPhone. Shot after shot, the iPhone pulled in more per-pixel detail, more accurate colors, and just blew me away with how sharp and noise-free the images were. To be clear, the Nikon was still shooting at F2.8 with great glass but the clear winner was the iPhone hands down. In this particular setting, Deep Fusion was a major factor in not only producing noise-free images but also in optimizing almost every aspect of the photo; lighting, color, and sharpness. Smart HDR did get an honorable mention in making some of the wide-angle shots noteworthy but the real hero by far was Deep Fusion. I’m amazed at what it can do, especially since this was just technically a beta version of the software!
Enough discussion let’s get serious with some side-by-side comparisons. Can you tell the difference? Which photos do you prefer?
Below are some additional portrait mode / depth-of-field comparison shots. It is somewhat unclear if Deep Fusion is enabled for portrait mode shots using the wide angle lens so take these with a bit more of a grain of salt.
Some quick last words. This is only the beginning of the computational photography revolution. Technologies such as Deep Fusion are only going to get better with time and I believe strongly we are on the cusp of seeing devices such as smartphones being able to produce images and video that is equal to, or perhaps even better than, complex dedicated general imaging devices such as DSLRs and Mirrorless cameras. This is great news for creative photographers as it will open doors to new ways to generate artistic imagery. Will dedicated cameras die? Not a chance, as there will always be some niche that requires specialized hardware. But given the results I have seen so far with the new iPhone 11 Pro; it would be a safe bet that there are many folks who will find this level of quality more than enough for their needs. And kudos to Apple for bringing two things I love together, machine learning and photography!