Facial Recognition & Annotation | Peakto

[@ednsb]

Ive been using it on some 200k images and parts of it are good finding images of people not looking at the camera directly or in low light or even facing away from the camera (not sure how you do that one) but also I am getting a fair number of wrong images. In one case peakto found 28k images of my daughter but some 10K were wrong. I've been working with your support team who has been quite responsive but here are a few things that are necessary. Firstly, I must be able to add a keyword for every face I am tracking. Secondly, the latest release took the ability to select, then go to another image contiguous to the first (like the first and last images in a row of five) and shift-click to select them all. Lastly, it would be great if they were more accurate. I know it is hard, especially when you have factors like my daughter's images because you are looking from infancy all the way to an adult. Plus she was a dancer for many years so we are talking about images that of dance performances.

[@CymePhotography]

Hi. Thanks for chatting. When you say 10k are wrong, are they wrong because they are pictures of someone who looks similar to your daughter ? Or are they completely off ? We will soon a video that goes in more details into more advanced options regarding faces. because the standard options are good in most cases but can get it wrong in instances where people look very similar. There is a way to create more and smaller clusters to ensure there is less confusion between similar people....and then merge those smaller clusters appropriately by giving the same name. 1.Regarding adding a keyword...I assume you want some sort of automation to do it ? because you can already do it. 2. You are right about this bug in the image view, but if I am correct, it is fixed in 2.0.7 The ability to create continuous or discontinuous selection is back in there. Note however, that selection of multiple faces in the face cluster view is not yet possible....we will add it in the near future 3. I don't know what kind of improvements we should make but everything you can send us to give us some samples on which we can work is very welcome. have a great daY.

[@ednsb]

@@CymePhotography Yes they were images that either had the face of someone other than my daughter but the image did have my daughter sometimes but also many images which weren't even images of people. As far as keywords there should be a way to add them from faces. I figured out I could select a collection as you are calling them then add the keyword of their name to all the images. I did that but when I went back to look at the images in On1 the keyword was not added. I am confused about your answer about continuous image selection. Are you saying it works within the main part of peakto but not in Faces? Because before I added the last 2 updates it was working and saved a lot of time. lastly, I need to understand a bit better how you define an image in faces. In the other apps that do this (and I will tell you right now Peakto was far superior in finding faces than Apple Photos or On1 and much faster but also almost too good as it was finding faces in images I would have to blow up to 4x to even see a face of a poster or painting) what I get is just the face. In Peakto, I got a lot of what I call failures where the face showing wasn't my daughter but she was in the image. I call that a failure but perhaps you are saying it is the image not the particular face you are finding?

[@arthuraizikovich2029]

Hi Join! Could you please make a video about the settings panel, where you would go in-depth about each setting how it works and when it does what it says it does? Also, could you tell me how do I know that catalogue has finished everything it needs for AI, indexing, faces etc? Thank you!

[@CymePhotography]

Hi. We will do it. Please note that we are thinking about redoing that panel in a more organised way because some screens expose important choices that do not get the explanations that would be required. Thanks for your request about clarity regarding task progress. You are right, some global tasks are split into parts and it is hard to know when everything is finished. On the other hand, there are a number of actions that do not require neither AI, nor Faces to have run ...for example, you can start viewing and annotating content while the ingestion is running. But we should improve the feedback we give when going to a screen or view that is incomplete because the process that populates it is still running or hasn't run yet. One way to give more info would be to expose a clear indicator next to a source that would say what process is running and what remains to be done. We will improve this... Thanks

[@ArnoldJ23]

Hi Matthieu. I have an issue where my watched folders won't sync properly. Run deep sync and run fast sync is greyed out. Run custom sync is available but doesnt do anything. My folder is on a nas which is connected via lan. I just updated to peakto 2.0.8., but no change so far. Only way to fix it is to re-ingest the entire folder again, which is not really a solution, since all the ai computing needs to be done all over again and the added metadata is also gone. Any idea how to fix this?

[@CymePhotography]

Hello Arnold, you can submit a ticket and let us know about your issue at (desk.cyme.io/portal/en/home). The development team will respond to you as soon as possible.