A7RII and Face Detection Auto Focus

This is an article on the focus characteristics of the Sony A7Rii. If you don’t own this camera you might want to skip this read.

This is the least desirable set with Face Detection. When S, M, L, Center or Zone focus areas are assigned, the Face Detection square stays GRAY and inactive when outside of the assigned focus area. If you take a picture the focus will be established in the chosen focus area, not the Face Detection box. The Face Detection box becomes WHITE and active only when the face detection box and the face of the subject is within the assigned focus area. The outline of the box begins as white, when focus is achieved, it turns green. Until you understand what is going on, it is easy to get out of focus images in this setup – you believe face detection is active but it is actually the focus area chosen. It results in faces out of focus.

There is an instance where this setup would be worthwhile. If I intended on taking many portraits with the face centered in the view finder I might use the Large (L) focus area slightly above center (to lock on the eyes). I would have no problems with this approach as long as I intended on centering the subject for each picture.

This is the best setup when faces are the subject of your photo. If the Wide Focus area is assigned then Face Detection is active with a white box throughout the viewfinder and turns green when focus is achieved. In AF-C Face Detection tracks the movement of the face. This setup is ideal as long as Face Detection identifies a face. The problem begins when a face is not detected, the AF system focuses and locks on a random area of high contrast, shown by green boxes or dancing ants. If the randomly picked focus area is not the one you want, then a menu change is required to “take charge” of focus. You could change to Manual focus or the S,M,L focus areas.

I assigned Eye-AF to the C3 button – a great added feature when the eye is prominent in the picture space.

Face Detection and Eye-AF – all of my testing has convinced me that Face Detection uses the same technology as Eye-AF and both systems seek to obtain focus of the eye. The only difference is when Eye-AF is used I know which eye will be in focus. This is useful when the subject is turned in a 2/3 position for example.

Sony could improve the system by combining systems. For example, when Face Detection is ON, if a face is detected it could automatically activate the wide focus area, AF-C and track the face in the viewfinder. Recomposing is easy. When a face is not present, or is missed by Face Detection because it’s in profile, or hidden behind hair, or eyeglasses then the system switches to AF-S and S (small) focus area. Then the user can choose the point of focus, lock and recompose. I would be able to choose the eyeglasses or the edge of the hair without pause. As it is now, the system randomly chooses areas of contrast and dancing ants takes over, then I delay shooting and give face detection a chance to work, if that doesn’t work, I give up and change settings.

Another habit of Face Detection worth noting is exposure metering. When face detection focuses on the face the exposure is adjusted to “properly” expose the face, even when in Multi-Metering Mode. The camera’s metering system switches over to something similar to “spot” metering. Of course there are advantages and disadvantages of this feature depending on your needs for image exposure. A face in shadow will result in added exposure and possibly blown highlights in the background. The better alternative may have been a multi-metering exposure and then raising the shadow slider in post-processing to bring the face in proper exposure.

Another option to face detection is Lock-on AF: Flexible Spot S (M or L) with AF-C the active mode. It a great feature because it doesn’t require a face to be detected yet it will track the lock on spot through the viewfinder. This is great for lock and recompose shooting of subject that might move. If the S (small) focus bracket isn’t grabbing the point you want then switch to M (medium). I have tested this a few times and have not been satisfied with the number of out-of-focus images.