chef 3810 Posted August 27, 2021 Posted August 27, 2021 31 minutes ago, Weevillo said: I only just found out about this project, is it usable at the current state and if so where would the dll file go? Hi! It would say it is beta. Are you on a Linux based machine, or are you running the most current beta of emby?
Weevillo 30 Posted August 27, 2021 Posted August 27, 2021 2 hours ago, chef said: Hi! It would say it is beta. Are you on a Linux based machine, or are you running the most current beta of emby? Right now I'm on windows and currently have not opted into the beta channel, although I wouldn't mind trying it out if there isn't too much of a risk. Though, by the sound of it it sounds like this is being tested on the linux platform, correct me if I'm wrong?
samuelqwe 317 Posted August 27, 2021 Posted August 27, 2021 (edited) 10 minutes ago, Weevillo said: Right now I'm on windows and currently have not opted into the beta channel, although I wouldn't mind trying it out if there isn't too much of a risk. Though, by the sound of it it sounds like this is being tested on the linux platform, correct me if I'm wrong? It’s just that only the beta of Emby has what you need to run this plugin on Windows. Linux builds of Emby already have these requirements. Anyways, the plugin correctly detects intro timestamps, but doesn’t do anything with them. There is no skip functionality. Edited August 27, 2021 by samuelqwe
Sammy 790 Posted August 28, 2021 Posted August 28, 2021 54 minutes ago, Weevillo said: Right now I'm on windows and currently have not opted into the beta channel, although I wouldn't mind trying it out if there isn't too much of a risk. Though, by the sound of it it sounds like this is being tested on the linux platform, correct me if I'm wrong? I run beta without issues mostly. "They mostly only come out at night, mostly." All kidding aside it runs fine and I usually wait a day or two before jumping on the next beta release to see how it is working out for everyone else. I have these two settings unchecked so I can control when I get beta releases downloaded and installed: 1
chef 3810 Posted August 28, 2021 Posted August 28, 2021 (edited) Save button is submit style Scrollable dialog fix "Has Intro" table cell is select box Removed Title Sequence Threshold input from UI (this can still be accessed in the configuration xml - default is 10) All Successful episode title sequence results are logged together in the task per season. (keep logging minimal during the calculations) IntroSkip_v2.0.2.4.rar Perhaps having one input for "maxDegreeOfParalellism" is needed? We can put it on the main page, and forget the dialog all together? This would set the amount of series, and episodes to calculate for both tasks in one input. Edited August 28, 2021 by chef 1
Sammy 790 Posted August 28, 2021 Posted August 28, 2021 rar not zip? BTW, I'm still waiting for the title sequence to finish from the last version! I'm 10 hours in now and it took nearly half a day to do the fingerprinting yesterday..
Sammy 790 Posted August 28, 2021 Posted August 28, 2021 If I do my math right, I've got 40 more hours to go, or more than 3 days?
chef 3810 Posted August 28, 2021 Posted August 28, 2021 52 minutes ago, Sammy said: If I do my math right, I've got 40 more hours to go, or more than 3 days? Oh boy. That is pretty long. I think there would be two log files in emby that would span the task time. One from yesterday, and one from this morning. Would I be able to see them? I know you have a large library, I'd love to make sure you're getting results.
Sammy 790 Posted August 28, 2021 Posted August 28, 2021 (edited) Here you go @chef.. embyserver 28082021_12am.txtembyserver 28082021_7am.txt Edited August 28, 2021 by Sammy
STR8 74 Posted August 28, 2021 Posted August 28, 2021 Don't feel bad @Sammy I'm at 51.7% and on my 3rd day... I know I have about 240 tv shows tho. So I'm thinking another 3 days and I'll be done lol!
Sammy 790 Posted August 28, 2021 Posted August 28, 2021 @chefwill have released two new versions by then making results irreverent maybe? 1
VictorPeeters 3 Posted August 28, 2021 Posted August 28, 2021 If I understand correctly, you can't use this to actually skip the intro's yet right?
samuelqwe 317 Posted August 28, 2021 Posted August 28, 2021 2 minutes ago, VictorPeeters said: If I understand correctly, you can't use this to actually skip the intro's yet right? Correct.
rbjtech 5284 Posted August 28, 2021 Posted August 28, 2021 (edited) @chef On the point above, (I know you have been rather busy lol) but did you get anywhere on potentially writing the Intro Start/End as extra chapter points in the emby chapter dB ? I guess as now there is a Intro database, there should be no reason why this can't be done as an optional task independent of anything else ? ie - lookup Item ID, retrieve existing chapter points, re-order with Intro Start/End, write new Chapters We can then 'use' the data in this fantastic plugin on a live emby instance - albeit manually via the chapters but now chapters are available in most of the clients, it brings this much closer. I'll lookup how to do it via the API .. edit - so getting the existing chapter points is easy enough, and the query Id matches your ResultId (I guess we should add SeriesID and SeasonID to speed up the query) but of course I have no way of accessing the IntrodB via the API.. btw - what is 'Confirmed' used for in the dB ? Edited August 28, 2021 by rbjtech 1
chef 3810 Posted August 28, 2021 Posted August 28, 2021 It shouldn't take that long to calculate the numbers. I had 1720 episodes complete in three hours running on an i5 processor. That was both task runtimes. That's 573 episodes an hour approximately... I will read Sammy's logs now and see what is happening.
Sammy 790 Posted August 30, 2021 Posted August 30, 2021 @chef IDK what happened but Episode Title Sequence must have restarted at some point. I did not monitor it as I was busy the rest of the weekend since I posted last. I'm going to kill it and install the latest version of introskip.dll embyserver 29082021_12am.txt embyserver 30082021_630am.txtembyserver 30082021_12am.txt
chef 3810 Posted August 30, 2021 Posted August 30, 2021 (edited) Oh boy, every plugin I have ever written seem to all break at the same time. My Alexa plugin went haywire and needed to be completely reworked yesterday, the auto organize upgrades started acting weird, and I had to go through it line by line to locate troublesome code. I have fixed them now. Holy mackerel! That took forever. Now back to this monster! Edited August 30, 2021 by chef 2
chef 3810 Posted August 30, 2021 Posted August 30, 2021 7 minutes ago, Sammy said: @chef IDK what happened but Episode Title Sequence must have restarted at some point. I did not monitor it as I was busy the rest of the weekend since I posted last. I'm going to kill it and install the latest version of introskip.dll embyserver 29082021_12am.txt 93.1 MB · 0 downloads embyserver 30082021_630am.txt 7.53 MB · 0 downloads embyserver 30082021_12am.txt 71.17 MB · 0 downloads Oh my gosh, last.fm just pollutes the log. I can't read it on my phone, I need notepad... One minute. 1
rbjtech 5284 Posted August 30, 2021 Posted August 30, 2021 (edited) Finally found some time to run this on a new Beta using my main Media libraries - the normal id scan is going to take a couple of hours (about half way through) but once complete, then I'm going to run the IntroSkip and do some analysis on the results - but given the size and complexity of the task on my ancient i5, I expect this will be 3-4 days away, maybe more .. Edited August 30, 2021 by rbjtech 1
chef 3810 Posted August 30, 2021 Posted August 30, 2021 Oh boy! Everything is broken... I literally have four IDE windows open right now. Thank goodness this is not my main profession. I would lose my mind . I don't know how the pros keep up.
rbjtech 5284 Posted August 31, 2021 Posted August 31, 2021 (edited) Hi @chef - So I tried to run the latest Introskip on my main TV shows collection and a few issues cropped up and I was forced to abandon it. At about 30% through the FP - the dB size for titlessequence.db was just over 1Gb .. After a graceful 'stop' - any db search in the Plugin now takes 30 seconds - presumably as the db size is large. Does it have any form of index to help with the query ? I also then tried to get some actual Sequence Detection completed - as expected, the logs were filled with 'missing FP' but on those it could find, detection went ok - and those details are showing ok in the Plugin (after 30 seconds.) However, I also had to abandon this as memory usage started to grow fast - I'm not sure if this was because of the missing FP's ? Part of me is thinking why are we splitting these items into two sets of 'complete' tasks - was there a reason for this ? Would it be better to FP a 'season or show', then run detection on that season, once detection is complete, then remove the FP data for completed shows (ie Status=Ended) and then repeat ? You are then left with a much more manageable database, quicker to search as it now only contains the FP data for 'Continuing' shows plus of course the Intro Start/End for all shows, which is, after all, the important bit. I think my system would 'get there' eventually, but if @Sammy is struggling with a 16core/32thread system, this process is going to kill most systems, especially people that run on low powered NAS devices for example - so perhaps a 'per season or per show' approach is needed to make it more manageable ? To end on a positive note - every detection it did get to was almost perfect - so the core detection technology works extremely well - it's just managing the journey to the end result that I feel needs a little more analysis .. edit .. So I've had a quick play with the dB - by removing all the data from the Fingerprint column (NULL) - now response in the Plugin are instant again - this is on a 10K row database (30% of my collection). After a vacuum, dB size is also back to under 1Mb. I'm going to restore the file and continue the FP task, but it does sound like the dB needs some indexes or maybe offloading to an archive dB if you do want to/need to keep the FP data. edit2 .. I've also just realised why the FP was eating so much CPU - I did not exclude the new Beta instance ffmpeg from my real-time AV scanning - doh! - so it was scanning every file it was trying to process.. Now ffmpeg is excluded, I can now run 6-8 parallel streams - each taking about 10% cpu.. Edited August 31, 2021 by rbjtech 2 1
rbjtech 5284 Posted September 1, 2021 Posted September 1, 2021 Update #2 So the FP process has finished - resulting db file is 2.5Gb in size .. I tried to run the Detection on my main machine but it just killed it with 100% memory usage (the VM only has 4Gb) - so I abandoned that and moved the Emby instance to my local machine and ran it from there instead on the previously FP's dB That is running - but embyserver.exe is currently using between 8-10 Gb of memory () - cpu is @ 50% - and this is with the default 4 threads. The logs show it is getting though the episodes, but extremely slowly .. but if it's having to do a lookup for each one (taking minutes each time) then this probably explains this. Are there any logs I can provide to help you identify the cause of the slow lookup and huge memory usage ? @Sammy Did you notice your memory usage during the Detection ? What size is your titlesequence.db file ? Thanks.
Recommended Posts