Finally Facial Recognition

By Mark

Lightroom has really become part of the whole Creative Cloud environment. With the release of LR4CC, Adobe has taken a huge step forward in improving performance as well as adding some great new features. This gives Roger and me a lot of material to cover in upcoming blogs. 

We have written extensively about the importance of keywording. One of the most requested features that LR has been missing is automatic facial recognition. Aperture and Picasa have had it for years. Reportedly, Adobe had been working on it for quite some time. As a first impression, well let’s just say they’ve left themselves plenty of room for improvement.

As part of the facial recognition feature, Adobe has added a new class of keywords called “Persons.” If you highlight a keyword and select “Edit Keyword” for an existing name you will now find a checkbox in the Keyword Properties page. Enabling it identifies the keyword as “Persons.”  

I have always tagged pictures with the names of the people in them. With this new feature turned on it identifies all of the people in the picture, even those I am not interested in tracking.  Luckily, you can select your entire People keyword list and make them all Persons as well.

How do you go about using it?  Well this is one of the few times I recommend going back into your folders because the process for identifying facial regions is very time consuming.  If you want to have it take on your entire catalog, do it before you go to sleep and it might be done in the morning.  I found it far easier to manage, one year at a time.  In the Library View, they have added a new People View.  

With a folder selected, if you click on it; it will open up a new People Pane.   

As you begin, LR will go through your images and identify all of the facial regions it can.  It will try and group them into similar looking individuals.  As you begin either typing in the name or dragging your already created keyword into the name box, LR will then start “learning” and will start recommending names for the unknowns.  A couple of things to note; First it is not always accurate in identifying all of the faces in a picture.  Some things like statues, definitely give it a hard time.  

This is not my talented musician friend Ric Sweeny.   

This however is, and you can see why it would potentially identify TJ as Ric.  

Secondly, you will notice that it missed Sarah entirely.   If it does not detect a face, you can just use the new Facial Region tool and draw a box around where you want it.

So, you can’t assume that it has found all of the people for you.  Thirdly, sometimes it identifies patterns in clothing or buildings as potential people.  You have to click on the box itself to make those regions go away. 

Back in the people pane, you can select multiple images it has correctly identified and then, either just click on the little check box confirming that it is correct or drag the pictures up into the Named People box.   If the person has not been previously identified, you can add a new name and it will create a new Person keyword.  It irritates me that LR does not let you set a default where the new name goes, it just adds it alphabetically.  Since all my people are in a top level hierarchy cleverly called People, I have to go and drag the names into the keyword list.

If you drag an image from the unnamed section and it is actually only part of a picture with multiple people, it will rename all of the people for you.  Definitely something which needs to be fixed. 

If you don’t care about who the person is, just delete the facial region and be on your merry way.   If you have a lot of images in your catalog, recognize that this will take a while to work through. Eventually you will wind up with all of the faces in the Named People pile and you will feel virtuous.  

However, just like in regular keywording, taking small bites will help you eat the entire elephant.   Adobe and of course KelbyOne have posted a lot of videos featuring the new capabilities.  We will be exploring them ourselves in the weeks to come.