Face Recognition Problem

[Mick Sulley]

Thanks for your reply Mario, I now understand!


On 25/09/17 14:19, Mario Frank wrote:
> Hey Mick,
>
> the purpose of the tags tab is to allow combinations.
> Consider you have an album "Family" which is full of family pictures and
> some junk album where also family pictures
> are located. You could then select the "Family" and junk album but then
> would also scan images from the junk album
> that are no family pictures. But if the family pictures in the junk
> album are tagged with the tag Family or some subtag
> of it, you can combine it in the speaking of:
> "Scan all pictures in the album Family and also all images with the tag
> Family"
> On the other hand, you can also just ignore the album tab and just scan
> in images having specific tags.
>
> Concerning the low detection quality: During Google Summer of Code,
> better algorithms have been introduced in digiKam
> but those are not yet integrated in the distributed version of digiKam
> as some testing has to be done.
> I would propose setting the detection parameter to at least 90 until the
> new algorithms are shipped.
> This is much slower but has less false positives.
>
> Regards,
> Mario
>
> Am 23.09.2017 um 20:07 schrieb Mick Sulley:
>> I seem to have found the answer.  I noticed that in Options under the
>> Albums tab there is a Tags tab which was set to Whole tags
>> collection.  I turned that off and now it just scans the specified
>> directory.
>>
>> What is the purpose of the Tags tab?
>>
>>
>> On 23/09/17 18:36, Mick Sulley wrote:
>>> Running DK 5.7.0 App Bundle.
>>>
>>> I have just run 'Scan collection for faces' with the settings Detect
>>> faces, and albums set to one specific directory.  It has scanned that
>>> directory and continued to scan the whole collection.  Is this a bug
>>> or have I missed something?
>>>
>>> Also I have the detection parameter set to 80, I guess the default as
>>> I have not changed it, and it is detecting loads of stuff as faces,
>>> clocks, curtains, bricks, all sorts of things.  I have had the odd
>>> thing in the past but nothing like this.  Do I need to change this
>>> setting?
>>>
>>> Thanks
>>>
>>> Mick
>>>