I have recently purchased aperture and was excited with all the features. I used faces on a couple of family portrait shoots which worked well.
However on returning from Egypt with around 1000 photos I am having difficulties. Ap is not recognising all the associated faces - in fact it suggests many more incorrect faces that the right ones - about 10 wrong faces to each correct one. I didn't experience this in when I used faces for the shoots. I have added a few thousand photos since but would be disappointed if this was the issue. Having used Picasa before I found never had issues recommending appropriate faces.
Can anyone help?
My workflow.
- import photos and keep originals in the pictures folder - not the aperture library
- rate them, name them, edit them
- choose faces in the inspector and I can see some faces of family at the bottom. I name a few
- I then click a face I have named to attempt to have aperture suggest appropriate faces
- Aperture thinks then displays several photos, most of which are incorrect.
- I then tried manually naming more of the same faces in the bottom section in order to see if that helps. It doesn't.
Again this all happened ok with my studio shoot but holiday pictures have not worked we'll at all - unlike Picasa which is excellent.
By the way - is there any way to have Aperture just use the current project for face recognition - I.e. it would be easier for me to only have faces from the holiday to choose from - not the entire library.
Thanks
I have been building quite a set of tires/wheels, and other classified objects for this, which is slowly helping toss similar objects into one area to speed the process (Another hobby is old, pre 1970’s Land-Rovers and British cars where the rims used were pretty consistent.) Now, why Aperture thinks the five bolt patters with central hub is a face is beyond me, but this oddity in the facial recognition software has resulted in finding some interesting “faces” in objects. To me, I treat this as an interesting feature.
Hi,
I mainly share the same experience as you. The face recognitionis quite a 1.0 feature. Tere is still alot of room for i provementa in the matching algorithm. My experience up to know is mainly thatmore face you have more confuse the system becomes….
So my workflow is usually the following:
1. Import the pictures with necessary common keyword and other global info
2. G quickly through the picture to identify the clear rejected one and flag them appropriately
3. Perform geotagging with map
4. Go theough all picture to place names
Then i will start my rating and real edit processs. Quiete painful but at least i am having more or less clean faces.