|
|
Hi open source photo friends,
I am currently giving digikam a try of having my huge photo collection
face tagged (talking > 2TB of photos) on the new 7.0.0 version built
from git.
What is kind of strange to me, is that both different options for people
detection (face detection, face recognition) are pretty low in resource
consumption on my machine.
The detect faces function in the first place is pretty much bound to
hard drive performance, I'd suppose, so it simply takes time, especially
given the amount of photos. For recognizing faces though, I have checked
"Work on all CPU cores" and it looks and feels like it is more at one
core only on a decent 1.5 load (8 cores, hyper-threaded). Looking at the
console log it is just pasting maybe 200 lines every 5-10 seconds and
not on a permanent basis too:
digikam.facesengine: Predicting face image
digikam.facesengine: m_threshold 0.7
digikam.facesengine: vecdata: 0.144905 0.0410241
Is that something everyone is experiencing or am I doing something
wrong? I'd like to have my machine working on full power to get the
stuff done. Same behavior I have seen at 6.4.0 too, though the new
FaceAI development is going to eliminate this.
BR;v
|
Locked
