Blog

[Jere_Jones]

This will be my "poor man's" blog. I'll just ramble about what's going on with SA. Feel free to comment, or just observe.

Last night I was up until 0200 working on rewinding. I think I have it right, but I'm still not happy with how rewinding affects the blocks measurement of edge matching.

Trying to make SA more "stream-friendly" than comskip is, I figure there could be multiple edge profiles in a single recording.

{side thought}
Actually, I should probably keep them instead of dumping them. Then I could see the relation from one part of the show to another. Interesting...
{/side thought}

To keep the logo boundries accurate, I need to rewind the decoder to the last positive identification when I first identify an edge profile or when I've gone a long time without matching an edge profile.

It is that rewinding that is a pain.

Jere

[Jere_Jones]

Interesting tidbit: When something goes wrong and a show gets split in two, there are a LOT of zeros in one of those files.

Note to self: MPEGs sometimes have more than 23 0 bits in a row.

UGH!!

Jere

[Jere_Jones]

Okay. I think I have all the possible crashes taken care of. As always, let me know if I missed something.

To take a break from the decoder, I've been working on the CC extraction. It's coming along nicely, but I don't know what format to output first: SMI? SRT? SUB? USF?

I don't use them, so I don't have a preference. Anyone?

Jere

[Jere_Jones]

Ok. SA can output SRT, SUB, and SMI. I even made sure that if SA is only extracting the CC's then it doesn't waste processor time decoding the audio and video.

If you want to know how much time is being spent decoding the audio and video, run it only extracting CC's and then run it detecting commercials and note the speed difference. For me, it's fairly large. I get 700fps detecting commercials and 1500fps for CC's only.

I need to work on that.

Jere

[Jere_Jones]

Audio is really frustrating me right now!!!!

Jere

[Jere_Jones]

I really hate MPEG audio.

Jere

[Jere_Jones]

I have determined that my audio decoder needs to be more flexible. I didn't program in the ability to change sampling rates. But, why on earth, would I need to? It's not something that should change. But, oh my I received a file that does have it change a lot!

AARRRGGGHHH!!!

Jere

[Jere_Jones]

I had duty Friday (that means that I show up to work at 6:30AM and leave at 9:00AM the next day, no sleep), so I spent a lot of time working on SA. The biggest problem, audio, was fixed.

After sleeping much of Saturday and working on SA for most of Sunday, I had 0.1.6 ready. It ran on all of my test files perfectly, even an HD snippet of the Pres talking. Commercials were correctly detected, no problems.

Then I released it.

I figured the code that displays the edge profile might be buggy. But it never crashed on me. Another ARRGHHH.

And, I must have broke something, because the crash reports are more than normal. But it works great on every show I have! Another ARRGHHH!

Anyway, back to the debugger...

Jere

[Jere_Jones]

Last night was fairly productive. Over the weekend, I updated my GUI library and the whole world fell apart. SA wasn't directly affected, but the problem meant I couldn't work on the GUI for SA's setup. ARRGGHHH!!

Anyway, last night, I found the problem. Who woulda figured? And the world continues to spin....

Jere

[Jere_Jones]

Ok. I thought I had this brilliant idea to improve the accuracy/usefulness of edgeprofiling. So I spent the last week completely rewriting that part of SA. Today I am getting frustrated because, while the preliminary testing looked good, sometimes it just goes off the deepend and locks onto something that is completely useless. ARRGGGHHH!!!!

Jere