P: Facial recognition and ISO Invariant

[Timothy.Spear]

My camera is fairly close to ISO Invariant. As such, I get better pictures by using Lr to enhance the exposure than use the camera to do the same and reduce the dynamic range.

However, I have found that facial recognition basically does not work with these images when exposure needs to be increased by two or three stops.

[Rikk Flohr: Photography]

Would you be willing to provide an example file or two for our engineers to review?

[Timothy.Spear]

@Rikk

Sure I will email it tonight, obviously Adobe cannot publish it anywhere. I have your email from previous discussions, subject will be "sample photo"

Tim

[Timothy.Spear]

@Rikk 

Email sent last night, please confirm receipt.

Tim

[Rikk Flohr: Photography]

Did not receive any emails from you. 

[Timothy.Spear]

@Rikk 

Resent with a link last night, instead of attached files.

[Rikk Flohr: Photography]

I am downloading them now. I will pass them along to Product Management. 

[Timothy.Spear]

I am finding more issues with facial recognition.

Two general patterns I am finding:

1. In high resolution photos, such as from my Sony A7RIV or from a 100mp Smartphone, a lot of people are missed. These high resolution images can be digitally zommed in a good by a factor of 20-30 times I am finding Lr misses.

2. Second related to ISO invariance. I often under expose by one to six steps and brighten in Lr. Faces are missed here.

I previously mentioned these issues here: https://community.adobe.com/t5/lightroom-ecosystem-cloud-based-ideas/lightroom-desktop-facial-recogn...

hoever the forum is not letting me append to the existing thread.

This issue is also getting worse as more family members provide me images from more and more phones with large sensors at family events to share the images.

Tim