randomguy wrote:
It seems exposure settings effect eye detect. It seems to detect eyes more easily if they are well exposed, and also the bigger in the frame the easier it seems. Unfortunately well exposed eyes sometimes may mean something else is blown. Once the eye is detected it seem to stick to it easier however. This is just couple quick observations trying it on a computer screen.
Yes, well the algorithm is properly reading from the live view stream which is also based on the jpeg settings. I wonder if setting a high d-range will boosted shadows could help. As long as we are shooting RAW, we can go nuts with those jpeg settings / picture style stuff.
