Subject Detection AF

Subject detection AF is an autofocus mode where the camera recognizes the type of subject in the frame and focuses on it without the photographer placing a focus point. A recognition model on the camera's processor scans the live image, finds a person, animal, bird, or vehicle, draws a box around it, and keeps focus on it as it moves. It appears on most mirrorless cameras released since about 2021.

Eye AF is one part of subject detection. Subject detection recognizes the whole subject first, such as a body, an animal, or a car. When the subject is a person or animal, the camera then narrows down to the face and finally the eye, which is the Eye AF stage. When no eye is visible, subject detection still holds focus on the head or body until the eye returns.

Common categories are people, animals such as dogs and cats, birds, and vehicles such as cars, motorcycles, trains, and aircraft. Some cameras add insects, and a few split planes and trains into their own types. The list depends on the brand, model, and firmware version. Manufacturers add new categories in firmware updates, so a camera bought with only human detection can gain bird detection later.

No. It needs a fast processor and an on-sensor autofocus system, so it is mainly found on mirrorless cameras released since around 2021. DSLRs offer face detection in live view at most. Smartphones detect faces and pets in their camera apps, which is a simple form of the same idea. Entry-level mirrorless bodies usually detect people only, while higher-end models add animals, birds, and vehicles.

AF-C, continuous autofocus, suits it best. Detection is designed to follow a subject, and AF-C keeps adjusting focus as it moves. In AF-S the camera detects the subject, focuses once when the shutter button is half-pressed, and then stops. That works for a posed portrait but wastes the feature on anything moving. Most cameras also allow detection in a tracking area mode that follows the subject anywhere in the frame.

Detection picks the most likely match, which can be the wrong one in a crowd, a flock, or a scene with a statue or poster that looks like a face. Auto subject mode adds another way to be wrong by guessing the category. Fixes include choosing the exact subject type, starting from a smaller focus area over the intended subject, or switching to a single point when detection keeps jumping.

It works, but less reliably. The recognition model needs enough light and contrast to identify a shape, so detection drops out earlier than basic autofocus does. Newer cameras with dedicated AI processors keep finding subjects at light levels where older models give up. When detection fails in dim scenes, a single focus point placed on a bright edge at the same distance is the fallback.

Detection identifies what the subject is and where it sits in the frame at that instant. Tracking follows a chosen subject from one frame to the next, using color, shape, and position. Current cameras combine them: detection finds the bird, tracking keeps the focus box on that same bird as it flies, and detection confirms the match each time the bird's shape changes.

Yes, and video is one of its biggest uses. The camera keeps a person or animal in focus as they move through the frame, which replaces a focus puller for solo shooters. Smooth tracking matters more than speed here, so many cameras have separate settings for how fast video focus reacts to a new subject. Detection also helps vloggers who cannot see the screen while recording.

Sometimes. The categories are broad, so bird mode often locks onto large insects and animal mode can catch a frog or a lizard, because the model responds to eyes and body shapes, not species. Results are less predictable than with a listed subject. When detection ignores an unusual subject, a normal single point or zone area still focuses on it in the usual way.