I shoot a lot of sports and face detection picks up way too many distant, out-of-focus faces in the crowd behind the subject. For instance, if you’re shooting a boxing match, there is rarely a need to apply face recovery to more than 2 faces in those photos.
I’m aware, thank you. I just think it would be easier to limit the number when applying to dozens/hundreds of images, rather than going into each photo and deselecting multiple images, even when the face recovery is limited to the auto-selected subject of the photo…it often selects way too many. It becomes quite time-consuming to manually deselect faces that are wayyy in the background. It may not be an issue for many photographers, but for professional sports photographers, there are often hundreds of people in the crowd behind the subject. See photo for example.
Currently you can only deselect faces and you cannot add faces so it may be difficult to select those faces you want … remember also that it selects low quality faces.
Thank you for the feedback. We recently released an update to the face recovery preferences that allowed limiting the selection to the subject mask, in addition to the “Auto” selection of low-quality faces. If I understand you correctly, you have tried both of these options and they both are selecting way too many small blurry faces by default. If you don’t mind, I would like to ask a few follow up questions.
When limiting the default face detection to the Subject mask, does your workflow ever involve refining or updating the auto detected mask? Or do you generally not bother (possibly because you do not use sharpening)? In either case, are there often small blurry faces that you would consider part of your subject, but don’t want to run face recovery on?
For the faces that are suggested by default that you do not want to run face recovery on, what is generally the reason? For example, is it (a) the results of the face recovery are bad or distorted, (b) face recovery produces a face that is too sharp relative to the blur the surrounds them, (c) the face recovery results are fine, but not worth the processing time to recover so many, or maybe another reason?
In regards to the suggestion of limiting the number of faces detected, do you imagine this being a global setting? In your workflow, is it common for you to have many images you are processing where you expect them all to have the same number of faces to be recovered (like many images from a single boxing match)? Or would you imagine it to be a per image setting that could easily be copied/applied to a large set of images?
If the limit was 2 for example, would your limit apply to the 2 largest faces, the 2 lowest quality faces, or the 2 highest quality faces? or maybe the 2 largest faces that were determined to be low-quality?
Thank you for taking the time to help improve the product. The basketball photo you shared is a great example of this issue. If you are able, would you mind sharing the original – and any other image you think would highlight the problem – via our Dropbox?