General questions / Suitable for large Collections?

[tilllt]

I downloaded and installed Photonix docker on my Debian 11 Homeserver Box. As a test case, i dropped in my ver unorderly digital photo collection and Photonix starts categorizing. Generally speaking, i am full of praise for the UI and the overall concept, but i think it has a long way ahead.

There are a couple of things I don't understand:

• there are a lot of photos missing from the catalog, sometimes Photonix only registers 2 in a series of 30 photos, without any apparent reason.

• looking at the output of docker-compose, i can see that it starts scanning at first and then stops "ingesting" but starts categorizing... Why is that, i would prefer to have ALL photos at first and then slowly get a finer grained access to them through the AI etc.

• my photo collection contains of a couple of 10.000 photos. I am not entire sure how much, but it's a lot. I don't think as is, Photonix can handle this amount of photos. The GUI fails to autoload the gallery after a certain point and generally things get extremely sluggish. I am not sure if Photonix was designed with that amount of photos in mind, but it would make the library more user friendly if there were some mechanisms of collapsing parts of it's maybe by year in the timeline view.

• on windows, on chrome / edge, the timeline view has no scrollbars, on Firefox it has scrollbars but the autoloading ends at some point although there are thousands of more photos in the timeline. Only browser that works somehow is Firefox ESR on my Debian server.

• generally speaking, the timeline view needs some info about the actual timeline. A timeline, pretty much by definition has some info attached to it, to see what moment in time you are looking it ATM, so maybe a year/month etc. Without this it's really difficult to look for photos from let's say 2019.

• it also would be nice to have a "jump to" feature. If you discover a photo using the categorizations, to have the option to jump to that moment in the timeline.

• i tried running Photonix on a different server with less RAM and there were a lot of process crashed in the log. On another server with more ram, it seems to run better, but i think there might be an issue with memory management.

• one last question: can a GPU be used to speed up the tensor flow processing? I have a AMD Radeon for realtime video transcoding in the server, so the GPU part is basically sitting unused and it would be great if I could put it to work for this task.

Thanks for Photonix, if it continues on this path it will be a great software some day!

P.s. the android app is unusable yet. When doing a quick scroll swipe over the collection, it opens a picture instead of scrolling down.

[legraml]

I fully agree with you! Photonix is aspiring but still buggy and sometimes even misleading in some cases. I can fully underline the importance of a "real" timeline and it needs also a lot more sorting and searching functions. Starting with browsing through selected categories or tags will lead in the GUI showing unselected photo sets once pressing the arrow keys.
The process of detecting new photos or whole sets remains a black box, there's no progress bar or other message function indicating the current status of the scanning process. There's no way to fire up a new scans.
The tagging is quite good compared to other solutions when it comes to object detection. But there's no way to correct false results for object detection.

[audioscavenger]

• can i ask you how you got passed the login spinner bug? i got the latest and cannot even login
• about the ram: yes i uses almost 2GB, i had to use a dedicated VM just for it

[tilllt]

Do you mean this?

#246 (comment)