“In focus” meant that the object you were pointing at was sharp for the majority of photography’s history. It’s fairly easy. To find the closest bright edge, a photographer could either manually adjust the lens or rely on a simple contrast-detection system. That was all. Even fifteen years ago, it would have seemed absurd to think that a camera could recognize a bird’s eye in midair and maintain its sharpness while wings blur and branches cut across the frame.
It no longer sounds absurd. More intriguing than a spec sheet upgrade is what has occurred over the last few years, primarily due to deep-learning AI integrated into camera processors. The meaning of “in focus” has subtly changed. It was a technical description once. Instead of just responding to contrast or distance, the camera now interprets what a photographer truly wants to be sharp, which is closer to an intention.
Jonas Classon, a bird photographer, put it simply: it used to take him ten years of practice and a ton of missed frames to get crisp action shots of Grey Owls. When Canon’s Eye Detection AF arrived, he was able to shoot owls in dense forest an hour after sunset, which would have previously been impossible. It’s not a slight improvement. That fundamentally alters the possibilities available to photographers.
Its engineering is actually quite intricate. When creating animal eye-tracking for their EOS iTR AFX system, Canon’s team had to deal with an issue that, when you think about it, sounds almost ridiculous: birds differ greatly from one another. There are very few similarities between a parakeet and a swan in terms of color, proportion, or silhouette.
A palm can hold a hummingbird. The distance between an albatross’s wings is three meters. To teach an AI to recognize all of them—resting, diving, flapping, and gliding—massive amounts of training images had to be gathered, and each misidentification had to be painstakingly corrected. During the system’s development, human hands were extended like birds and flowers were flagged. The engineers worked on it iteratively for several months at a bird sanctuary in Shizuoka Prefecture, southwest of Tokyo.
Sony approached the issue from a slightly different perspective, emphasizing perseverance and speed. As the underlying AI developed, their Real-time Eye AF system—which was commended for being nearly aggressively accurate in fast-action scenarios—was expanded from human subjects to animals. The end product is a system that can use movement, partial obstruction, and shifting light to maintain lock on a pet’s eye, something that previously required a highly skilled human operator to even approximate.
However, it’s important to be truthful about the boundaries. Although impressive, these systems are not perfect. A camera may focus on the incorrect animal in crowded wildlife scenes, especially if the animal is better lit or facing the lens more directly.
Even the best systems can stop tracking in the middle of a sequence when there are obstacles like grass or branches cutting across the subject in low light. Photographers with strong compositional instincts occasionally bring up a more subtle point: the AI determines what is important. Having the camera override your judgment, no matter how well-meaning, can seem like a minor annoyance when you’ve spent years honing your eye for precisely where focus should land in a frame.
However, the practical implications of this change are difficult to overestimate. On challenging moving subjects, hit rates that were previously at twenty or thirty percent are now getting close to ninety or higher. Photographers are regularly taking pictures that were just not possible in the past; this isn’t because their abilities have improved, but rather because the tool’s capabilities have been radically altered.
This is part of a larger pattern that goes beyond cameras. RETINA is a deep-learning algorithm created by researchers at the University of Maryland. It uses raw eye movement data to predict human decision-making within seconds of visual exposure, long before a person makes a conscious decision. The same fundamental reasoning holds true: AI’s ability to track eye movements and deduce meaning from them is developing into a capability that spans a variety of industries, including consumer behavior, photography, and medical diagnostics.
The practical outcome is still being worked out, particularly for wildlife photographers. Two or three sharp frames out of ten were once produced by Jonas Classon’s great crested grebes, which move their heads quickly from side to side during mating dances. The hit rate is now almost flawless. That is not a tale of technology taking the place of expertise. It’s a tale of skill at last having a tool that can match it.
In 2026, the definition of “in focus” is more expansive than it was previously. It indicates that the camera recognized your intent.
