Author Topic: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread  (Read 942647 times)

Offline JohnKiel

  • Member
  • Posts: 15
  • United States
  • Liked: 10
  • Likes Given: 0
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1300 on: 06/02/2014 11:17 PM »
The online editor, http://spacex2.slapbet.org/, now has buttons to allow loading MMBs directly from spreadsheets that are now listed in the wiki.

These spreadsheets probably still need some cleaning up and updating to sync them up with the wiki tables.  I believe myroslav is working on embedding images in the sheet.

Make sure that MMBs entered in the sheets don't have any notes or the like in them.

Thanks to myroslav, a.k.a. lgjy98d, for setting the sheets up, and IainCole for the online editor and his help in getting these changes to the editor live.

The online editor, http://spacex2.slapbet.org/, now has buttons to allow loading MMBs directly from spreadsheets that are now listed in the wiki.

These spreadsheets probably still need some cleaning up and updating to sync them up with the wiki tables.  I believe myroslav is working on embedding images in the sheet.

Make sure that MMBs entered in the sheets don't have any notes or the like in them.

Thanks to myroslav, a.k.a. lgjy98d, for setting the sheets up, and IainCole for the online editor and his help in getting these changes to the editor live.

The spreadsheet for iframe 8 is the one that we have been working on as a template. It has the images in etc.

Offline DecoLV

  • Full Member
  • *
  • Posts: 137
  • Boston, MA
  • Liked: 80
  • Likes Given: 23
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1302 on: 06/03/2014 12:52 AM »
1 - Do you think we need better tutorials for using the editor? What would you consider better?
2 - Do you think the online editor should be more intuitive to use? What is not intuitive right now?
3 - Do you need more feedback flipping single bits or changing bitpositions? Do you need to see the actual bits? What kind of feedback apart from the resulting frame?
4 - Do you need to have a better explanation about MPEG4? Do you want to see a concrete example based on the actual video data?
5 - Do you want simpler and more concrete tasks to do? It's now divided into roughly 300 parts. Should we divide it more? Or should it be divided differently?

Quick feedback on this....

No experience. I went through the p-frame tutorial for about 20 minutes, and sort of got it a little, but not enough to be clear what I was doing. I don't think anything is wrong with the tutorial, but even if I invested  a couple of hours, I doubt I could make a meaningful contribution. I would just feel in the way and too scared to commit anything to the wiki. I appreciate you folks inviting us in, but I truly don't think this is a job for noobs....you have to know what you are doing, a little. Sorry. You folks are awesome and your work is appreciated. Good luck!


Offline Untribium

  • Member
  • Posts: 32
  • Switzerland
  • Liked: 32
  • Likes Given: 78
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1303 on: 06/03/2014 01:16 AM »
1 - Do you think we need better tutorials for using the editor? What would you consider better?
2 - Do you think the online editor should be more intuitive to use? What is not intuitive right now?
3 - Do you need more feedback flipping single bits or changing bitpositions? Do you need to see the actual bits? What kind of feedback apart from the resulting frame?
4 - Do you need to have a better explanation about MPEG4? Do you want to see a concrete example based on the actual video data?
5 - Do you want simpler and more concrete tasks to do? It's now divided into roughly 300 parts. Should we divide it more? Or should it be divided differently?

Quick feedback on this....

No experience. I went through the p-frame tutorial for about 20 minutes, and sort of got it a little, but not enough to be clear what I was doing. I don't think anything is wrong with the tutorial, but even if I invested  a couple of hours, I doubt I could make a meaningful contribution. I would just feel in the way and too scared to commit anything to the wiki. I appreciate you folks inviting us in, but I truly don't think this is a job for noobs....you have to know what you are doing, a little. Sorry. You folks are awesome and your work is appreciated. Good luck!

First of all, thanks for sharing, this really helps! The thing is, I'm a noob myself :D I mean, I study computer science, but I've never done anything even remotely close to what we're doing here. Also. one of my least favorite topics was visual computing :)
One problem that I can see, is that it's incredibly difficult if the first thing you work on is a badly corrupted p-frame. Most of us started on frames like i-frame 8 (back then, now 10), which was almost in perfect shape, and you could quickly make progress. Then there was i-frame 2 (now 4), which was a little bit tougher, but still doable, then 9 (now 11)...you get the point :) Starting on a bad p-frame is like trying to teach integral calculus in first grade - you wouldn't understand anything cause you need to practice simpler stuff first. What might help is a step by step tutorial how to recover e.g. i-frame 4, so people could go through it one part at a time and learn the different techniques required to do other stuff.
Another "problem" is that some people here have gotten incredibly efficient, so one might argue that it's better to leave it to them. I don't think so, though. I find that I spend the most time trying a range of maybe 100 to 400 bit positions for a specific block. So I guess one way to make it easier for people to join would be to post the range and then people can post the most promising results back here.
It all comes back to this website where we generate thousands of images and have people vote on them :D It sounds like a lot of fun, but I'm not sure how well it would work...Anyways, gotta work on 101 some more :)
« Last Edit: 06/03/2014 01:17 AM by Untribium »

Offline Robotbeat

  • Senior Member
  • *****
  • Posts: 28049
  • Minnesota
  • Liked: 7861
  • Likes Given: 5233
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1304 on: 06/03/2014 01:53 AM »
I have a suggestion: a basic tutorial video where you fix something and explain everything you're doing and why. Just an idea.
Chris  Whoever loves correction loves knowledge, but he who hates reproof is stupid.

To the maximum extent practicable, the Federal Government shall plan missions to accommodate the space transportation services capabilities of United States commercial providers. US law http://goo.gl/YZYNt0

Offline meekGee

  • Senior Member
  • *****
  • Posts: 8179
  • N. California
  • Liked: 4420
  • Likes Given: 868
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1305 on: 06/03/2014 03:40 AM »
I have a suggestion: a basic tutorial video where you fix something and explain everything you're doing and why. Just an idea.

Yeah, but scramble it so you can only make out a facial expression twice throughout, and lots of green please.
ABCD - Always Be Counting Down

Offline wronkiew

  • Full Member
  • *
  • Posts: 186
  • 34.502327, -116.971697
  • Liked: 105
  • Likes Given: 121
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1306 on: 06/03/2014 05:05 AM »
I finished putting together a clock fix editor, so anyone who wants to help out can give it a try. Find it here:

http://adama.nocdirect.com/~wronkiew/spx_crs3/clockfix_edit.php

It's quite limited compared to IainCole's amazing mmb editor, but it does let you write clockfix instructions and test them out. Also it shows any potential parsing errors. It has no image inspection tools, so you will need to download the displayed frame and open it in a program like GraphicConverter or Photoshop whatever the Windows equivalent is. It will also be very helpful to download the isolated p-frame images from the wiki.

If you want to try it out, select frame 101 and copy in these instructions:

# erase (row,begin,end)
464,32,40
465,32,39
466,32,39
467,32,39
468,32,39
469,32,39
470,32,39
471,32,39
472,32,39
473,32,39
474,32,39
475,32,39
476,32,39
477,32,39
478,32,39
479,32,39
# digit (char,x,1even/2odd/3both)
9,32,3
:,46,3

Then press the Submit button to see the fixed frame.

The editor/tester does not save the clockfix instructions for you. Once you have cleaned up the gibberish and repainted the digits, save the text to a file clockfix-nnn.csv with the frame number, for example clockfix-047.csv. Then upload it to the wiki.

As I mentioned on the wiki page, only repaint digits if you are 99.9% confident of what they are. The intent is to make numbers readable, not to interpolate between "good" timestamps.

I would welcome any feedback on making the editor more usable.

Offline deruch

  • Senior Member
  • *****
  • Posts: 2188
  • California
  • Liked: 1695
  • Likes Given: 3400
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1307 on: 06/03/2014 06:49 AM »
I finished putting together a clock fix editor, so anyone who wants to help out can give it a try. Find it here:

http://adama.nocdirect.com/~wronkiew/spx_crs3/clockfix_edit.php

It's quite limited compared to IainCole's amazing mmb editor, but it does let you write clockfix instructions and test them out. Also it shows any potential parsing errors. It has no image inspection tools, so you will need to download the displayed frame and open it in a program like GraphicConverter or Photoshop whatever the Windows equivalent is. It will also be very helpful to download the isolated p-frame images from the wiki.

/snip/

As I mentioned on the wiki page, only repaint digits if you are 99.9% confident of what they are. The intent is to make numbers readable, not to interpolate between "good" timestamps.

I would welcome any feedback on making the editor more usable.

Do you need to delete?  For many of the frames, the numbers are simply missing.  Can we just edit in the correct number without deleting the background?  e.g. for frame #201 giving it this code:

# digit (char,x,even/odd/both)
1,18,3
9,32,3
:,46,3
3,58,3
Shouldn't reality posts be in "Advanced concepts"?  --Nomadd

Offline wronkiew

  • Full Member
  • *
  • Posts: 186
  • 34.502327, -116.971697
  • Liked: 105
  • Likes Given: 121
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1308 on: 06/03/2014 07:13 AM »
I finished putting together a clock fix editor, so anyone who wants to help out can give it a try. Find it here:

http://adama.nocdirect.com/~wronkiew/spx_crs3/clockfix_edit.php

It's quite limited compared to IainCole's amazing mmb editor, but it does let you write clockfix instructions and test them out. Also it shows any potential parsing errors. It has no image inspection tools, so you will need to download the displayed frame and open it in a program like GraphicConverter or Photoshop whatever the Windows equivalent is. It will also be very helpful to download the isolated p-frame images from the wiki.

/snip/

As I mentioned on the wiki page, only repaint digits if you are 99.9% confident of what they are. The intent is to make numbers readable, not to interpolate between "good" timestamps.

I would welcome any feedback on making the editor more usable.

Do you need to delete?  For many of the frames, the numbers are simply missing.  Can we just edit in the correct number without deleting the background?  e.g. for frame #201 giving it this code:

# digit (char,x,even/odd/both)
1,18,3
9,32,3
:,46,3
3,58,3

It'd be best to get rid of the squiggly line between the 3 and the 5.

Offline Quialiss

  • Member
  • Posts: 75
  • Liked: 82
  • Likes Given: 29
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1309 on: 06/03/2014 10:10 AM »
Quick note to our MB position finding wizards:  It is definitely worthwhile to go looking at the places data was added in frames 6, 7, 8, and 12.  The relevant bitpositions and number of added bits are on their respective wiki pages. 

I just found a block of 12 macroblocks in frame 7.  I think that's actually the first i frame data I've recovered.  :D 

Offline SwissCheese

  • Full Member
  • *
  • Posts: 164
  • Liked: 249
  • Likes Given: 90
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1310 on: 06/03/2014 10:15 AM »

1 - Do you think we need better tutorials for using the editor? What would you consider better?
2 - Do you think the online editor should be more intuitive to use? What is not intuitive right now?
3 - Do you need more feedback flipping single bits or changing bitpositions? Do you need to see the actual bits? What kind of feedback apart from the resulting frame?
4 - Do you need to have a better explanation about MPEG4? Do you want to see a concrete example based on the actual video data?
5 - Do you want simpler and more concrete tasks to do? It's now divided into roughly 300 parts. Should we divide it more? Or should it be divided differently?

Can you please at least sort the above numbers in order of priority? And then explain why you think some are more important than others? And maybe then try to answer (some of) the questions? Thanks!

Could you also tell what you read (tutorial/wiki page wise) before using the online editor? Because maybe we can do something about the (more prominent) placement of the explanations already available on the wiki/forum.

I really like to know. And hopefully you can help us here.

Thanks in advance.

Regards,

arnezami

Just a few remarks not exactly related to this, but that should help people understanding how to clean p-frames:
- Enhance the contrast with X:59:C0 / X:59:80 / X:59:60 / X:59:40 (usually C0 is the best, but sometimes another is better), it helps a lot seeing errors and good sequences. Remove this when you're finished.
- Correct MB look somehow "natural", they usually don't have crosses in them, have information in the full MB, don't have funny colors.... Try with an already corrected p-frame to understand better what was removed.
- Align features with the previous and following p-frame (surf on water works very well).
- Bitflips work quite well for i-frames; they usually don't work for p-frames: even when it seems to solve your issue, when you look carefully at the MB around your correction you find some "unnatural" features. It is really rare that it really corrects an error.

Otherwise, the more you do it, the better you get, and when you get back to earlier work, you see that you made many mistakes :)

« Last Edit: 06/03/2014 10:16 AM by SwissCheese »

Offline SwissCheese

  • Full Member
  • *
  • Posts: 164
  • Liked: 249
  • Likes Given: 90
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1311 on: 06/03/2014 11:02 AM »
The online editor, http://spacex2.slapbet.org/, now has buttons to allow loading MMBs directly from spreadsheets that are now listed in the wiki.

These spreadsheets probably still need some cleaning up and updating to sync them up with the wiki tables.  I believe myroslav is working on embedding images in the sheet.

Make sure that MMBs entered in the sheets don't have any notes or the like in them.

Thanks to myroslav, a.k.a. lgjy98d, for setting the sheets up, and IainCole for the online editor and his help in getting these changes to the editor live.

The spreadsheet for iframe 8 is the one that we have been working on as a template. It has the images in etc.

Just updated the mmb's of this part, the images come automatically, really nicely done!

Offline Ford Mustang

  • Live Coverage Editor
  • Global Moderator
  • Senior Member
  • *****
  • Posts: 12192
  • Virginia
    • Astro95Media
  • Liked: 69
  • Likes Given: 28
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1312 on: 06/03/2014 11:52 AM »
I must've been living under a rock for the past month and missed this thread.  Just wanted to say, you guys have done an absolutely amazing job pulling all this data out of the raw TS file.  Keep up the great work, it's very impressive to know just how diverse this forum is in everything!

Offline JohnKiel

  • Member
  • Posts: 15
  • United States
  • Liked: 10
  • Likes Given: 0
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1313 on: 06/03/2014 12:11 PM »
The online editor, http://spacex2.slapbet.org/, now has buttons to allow loading MMBs directly from spreadsheets that are now listed in the wiki.

These spreadsheets probably still need some cleaning up and updating to sync them up with the wiki tables.  I believe myroslav is working on embedding images in the sheet.

Make sure that MMBs entered in the sheets don't have any notes or the like in them.

Thanks to myroslav, a.k.a. lgjy98d, for setting the sheets up, and IainCole for the online editor and his help in getting these changes to the editor live.

The spreadsheet for iframe 8 is the one that we have been working on as a template. It has the images in etc.

Just updated the mmb's of this part, the images come automatically, really nicely done!

Unfortunately, it looks like the MMBs can create an image request URL that's larger than the ~2KB Google Sheet's IMAGE function can manage.
« Last Edit: 06/03/2014 12:14 PM by JohnKiel »

Offline saliva_sweet

  • Full Member
  • ****
  • Posts: 552
  • Liked: 422
  • Likes Given: 1325
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1314 on: 06/03/2014 12:39 PM »
- Enhance the contrast with X:59:C0 / X:59:80 / X:59:60 / X:59:40 (usually C0 is the best, but sometimes another is better), it helps a lot seeing errors and good sequences. Remove this when you're finished.
This can usually be cranked even further by X:58:80, X:58:C0 or X:58:E0

Offline Lourens

  • Full Member
  • *
  • Posts: 156
  • The Netherlands
  • Liked: 206
  • Likes Given: 304
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1315 on: 06/03/2014 12:45 PM »
These would be the two resources I kept going back to for reference.  I STILL don't understand why the propagation of values changes so dramatically sometimes, but I've learned to work with it.  (If I understand correctly the direction a block inherits from is dynamic.. I just don't quite understand why it switches seemingly arbitrarily due to luminance changes)
Which direction to use is determined by the DC values of the left, top-left and top neighbours (section 7.4.3.1 in the standard). If the DC values of the left and top-left neighbours are closer together than the DC values of the top-left and top neighbours, then MPEG4 concludes that the image changes less vertically than horizontally, and uses the top neighbour as a reference. If the top-left and top neighbours are closer in DC values, then it will use the left neighbour to base the current block on. So you need to take into account changes in those three neighbours. Does that make sense?

3 - Yes, absolutely! Actually seeing the bits could be enormously helpful.
However: When I started trying to correct chroma errors by flipping bits, I got confused by how often flips would change the length of the block, because I was expecting some kind of prefix based encoding i.e. 2 size bits encoding n and then n payload bits. Turns out, they're using variable length coding, which makes a ton of sense, but it also makes it more difficult to actually interpret the bits you're seeing (because they essentially don't have a meaning anymore). So what we need in order to be able to find the flipped bit, is the encoding table (I'm guessing it's some kind of tree...huffman comes to mind :) ). I'm guessing there's one table per frame (maybe only iframe?...), so it would be easy do decode them all once (they shouldn't change, right?) and post them on the wiki, maybe even make them available on the online editor. Again, just an idea :)
It's even easier, the tables are fixed by the standard. The tables in Appendix B contain all the secret codes you'll need :).

Offline Untribium

  • Member
  • Posts: 32
  • Switzerland
  • Liked: 32
  • Likes Given: 78
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1316 on: 06/03/2014 01:07 PM »
-snip-
3 - Yes, absolutely! Actually seeing the bits could be enormously helpful.
However: When I started trying to correct chroma errors by flipping bits, I got confused by how often flips would change the length of the block, because I was expecting some kind of prefix based encoding i.e. 2 size bits encoding n and then n payload bits. Turns out, they're using variable length coding, which makes a ton of sense, but it also makes it more difficult to actually interpret the bits you're seeing (because they essentially don't have a meaning anymore). So what we need in order to be able to find the flipped bit, is the encoding table (I'm guessing it's some kind of tree...huffman comes to mind :) ). I'm guessing there's one table per frame (maybe only iframe?...), so it would be easy do decode them all once (they shouldn't change, right?) and post them on the wiki, maybe even make them available on the online editor. Again, just an idea :)
It's even easier, the tables are fixed by the standard. The tables in Appendix B contain all the secret codes you'll need :).

Oh...that seems kinda silly...I mean, the whole point of VLC is to optimize the tables to assign the shortest possible codes to the most common values in the chunk of data you're encoding, right? Sure, you can assume that small numbers will generally be more common, but I feel like a prefix based encoding should be more efficient for this task (it's probably not, I think these people know more about this than I do ;) ). Oh well, I'm definitely not complaining, should make things easier. Thanks for looking it up :))

Offline Quialiss

  • Member
  • Posts: 75
  • Liked: 82
  • Likes Given: 29
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1317 on: 06/03/2014 01:13 PM »
If the DC values of the left and top-left neighbours are closer together than the DC values of the top-left and top neighbours, then MPEG4 concludes that the image changes less vertically than horizontally, and uses the top neighbour as a reference. If the top-left and top neighbours are closer in DC values, then it will use the left neighbour to base the current block on.

Aaah, this is the magical part I was missing.  Thank you! 

Offline Quialiss

  • Member
  • Posts: 75
  • Liked: 82
  • Likes Given: 29
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1318 on: 06/03/2014 02:14 PM »
Unfortunately, it looks like the MMBs can create an image request URL that's larger than the ~2KB Google Sheet's IMAGE function can manage.

A bit ironic considering the editor had to be modified to deal with the long MMBs.  It still works for all the individual p frames though, which is great! 

Offline JohnKiel

  • Member
  • Posts: 15
  • United States
  • Liked: 10
  • Likes Given: 0
Re: SpaceX Falcon 9 v1.1 CRS-3 Splashdown Video Repair Task Thread
« Reply #1319 on: 06/03/2014 03:04 PM »
Unfortunately, it looks like the MMBs can create an image request URL that's larger than the ~2KB Google Sheet's IMAGE function can manage.

A bit ironic considering the editor had to be modified to deal with the long MMBs.  It still works for all the individual p frames though, which is great!

I'm tinkering with a work-around.  Basically a proxy that allows a much shorter URL to request the image.  Currently sorting out how to cache and respond with 304's to avoid quickly using up the paltry 10GB/month I get on my chosen host. (If anyone with more bandwidth is willing to host the simple php image proxy script once I'm done, let me know.)

Tags: