<html><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8"></head><body><div>In your case I would started with the default accuracy settings (1/2) and if keep getting plants and unwanted faces would moved the slider to 1/4 and reviewed the results</div><div><br></div><div><br></div><div><br></div><div id="composer_signature"><div style="font-size:85%;color:#575757" dir="auto">Sent from my Samsung Galaxy smartphone.</div></div><div><br></div><div style="font-size:100%;color:#000000"><!-- originalMessage --><div>-------- Original message --------</div><div>From: Boudewijn <wankelwankel@yahoo.com> </div><div>Date: 2017-02-28 3:48 PM (GMT-07:00) </div><div>To: digiKam - Home Manage your photographs as a professional with the power of open source <digikam-users@kde.org> </div><div>Subject: how to optimally use face*tion </div><div><br></div></div>Hi List,<br><br>TL;DR: how do I make optimal use of face detection/recognition? I seek to <br>improve recognition results.<br><br>I use DK, but not as a professional. Mostly family images. I am happy with the <br>face detection because it helps to identify and tag pictures with friends and <br>family on them quick and easy. <br><br>I think I do not use the most optimal way, so I am hoping for feedback. <br><br>Detecting faces, there are 4 kinds of false hits, and one that I want to tag:<br>1. detected as face, but it is a plant <br>2. detected as a face, but it is a painting of a face<br>3. detected as a face, but someone in the crowd I don't know<br>4. detected as a face, but I am not interested in tagging this specific <br>instance of the face of this acquaintance<br>5. detected as a face that I want to tag <br><br>One last category, I'm not sure where to put it: there is a face in the middle <br>of the thumbnail, but it is much less zoomed into than usual and got other <br>things around it (faces, plants, etc). <br><br>How do I help/disturb the training data most for <br>1. detection<br>2. recognition<br><br>1. I would guess that detection is helped by pruning the non-face false hits, <br>and keeping faces I am not interested in. I use the upper-right X on the face <br>thumbnail to reject it, and the lower-right 'no parking' symbol (circle with <br>diagonal slash) to make the thumbnail disappear from the unknown detections.<br><br>2. I would guess that recognition is helped by having a small number of <br>disperate face tags to choose between, and a large number of confirmed faces <br>with each name. <br><br>Lacking time to hang as many tags on each image as I would like, I kept, in <br>the end, to a minimum of tagging people (the old fashioned way) and places <br>(hardly geo-tagged) as a minimum for making fast intersections between <br>combinations of people and visited places (fully realizing those would be the <br>first to be automated and making it futile to spend time on those...)<br><br>So far face detection helps me, but face recognition does not yield the <br>results I hoped for so I end up going through the set of images more than once <br>to check for false negative names, false positive names and just for fun.<br><br>Maybe I should mention that making the process do anything at all is no <br>problem at all after selecting some albums or tags, moving the parameters left <br>or right and selecting or deselecting the "All cpu's are belong to DK - all <br>cpu's - all cpu's are belong - .. sorry, couldn't leave it ;-) ). <br><br>Thanks for reading through all of this. Thanks even more for everyone involved <br>in bringing digiKam!<br><br>Best regards,<br><br>Boudewijn<br><br>PS: should the parameter moved to "high accuracy", like 95%, (only) make fewer <br>false positive detections, or does it (also) increase false negatives? And <br>does it make training data into garbage if I run it in low accuracy once? In <br>other words: fast will give me (only) results that are faces for sure after a <br>quick scan (skipping difficult faces and fetching plants), and accurate will <br>give me "hardly any" plants and "will not" skip difficult faces? <br>Will this improve after training, or is each image scrutinized on its own? <br></body></html>