Usability of face tagging has some weaknesses
I'm new to Peakto 2.4 and in the process of face tagging. First, the recognition quality is often surprisingly good – congratulations.
Second, usability and handling is (in my eyes) overly complicated and partly surprisingly poor, sorry to say so. Am I correct by assuming nobody of the dev team has experience with Aperture's faces? Apple put in some pretty amazing usability there. Peakto's usability is often the opposite of amazing. Or – maybe worse – there is a better way of handling, but well hidden and not easy or intuitively to be discovered?
Maybe I just need to learn much more, but I'll try to describe the things which I find annoying:
- Change of focus in the window view: Why make things soooo complicated?
- in the window view of "untagged faces" I often see clusters of a couple of faces of different persons I can tag with a name.
- Selecting one cluster, checking if it's content is correct (while remaining in overview window) and tagging the name to the cluster, changes immediately the whole view.
- Instead being able to tag the other familiar faces, I get a couple of new clusters. The content of the window is constantly changing and creates more stress than necessary. I get angry because of this poor usability.
- Tagging multiple clusters at once – how?
- in the window view of "untagged faces" I often see clusters of 2...10 images. Sometimes, depending on thumbnail size, three or more clusters are of the same person.
- How can I select all clusters belonging to one face (after checking the cluster has only correct positives) and apply one tag for all selected clusters?
- I had to learn the first thumbnail of a cluster can be misleading very much.
- Often it really contains the faces of one single person - great.
- In some cases it's terribly wrong and contains dozens, few times hundreds of profile faces from awkward angles.
- If I then "believe" the cluster thumbnail and don't check the cluster content first, it becomes very time consuming to clean out the "false positives" of a cluster.
- Sometimes it's easier to mark the "correct" faces and tag a new name to them, then apply a different tag name to the false positives and rename the old face cluster.
- But Peakto is rather slow and hesitant to get the idea, so this workaround might have success or bad luck.
- Low quality of a face image – can I adjust the level of recognition?
- It's great that Peakto can detect faces far in the background, drowning in bokeh swamp. Unfortunately pretty useless.
- It's no point in tagging blurred faces. Can I set up a default minimum quality somewhere? Currently there's "Strict clustering" activated, but I don't reckon this to be a limit for image quality?