Tragic Lantern for EOS M

Started by coutts, April 17, 2013, 01:43:28 AM

Previous topic - Next topic

0 Members and 6 Guests are viewing this topic.

gary2013

Hi Max,
I like this newer version being an exe. I tried it on one raw file from The M and it seems to work okay. It does hang at the start showing only a black screen with a blinking tiny white box at the top right. It took about 25 seconds before it started showing action. I almost gave up thinking it was not going to work. Can you make it so some text pops up saying "Processing now" or something similar so people know it's working and that we should be patient.

Gary

maxotics

Thanks Gary!  Good idea.  That wait you're seeing is the program creating a copy of the RAW file to work with.  Anyway, put a message in.  I noticed that on some really large files the program dies.  Don't know what causes it, if it will happen again, we'll have to see.

The program could definitely be speeded up by a master C# person.  Not a priority for me at moment.

On the Black Magic Pocket Cinema Cam

The camera is a joy to use after coming from the buggy build of ML on the EOSm.  You turn it on, hit iris, then focus, then shoot.  ProRes goes straight into the NLE.  the BMPCC is light and small.  But of course, it is $1,000.

What I didn't expect is the moire and aliasing caused by the sharpness of the image.  So the image needs to be softened a little bit.

The good news is that the 720p crop-mode image from the EOS-M is just as good to me, dynamic range wise, as the BMPCC.  If you want a nice grainy 16mm film look, I don't think any cameras, even the black magics, are going to beat the EOS-M or t4i, etc.  You may not believe me, but 720p is enough for me.  It's the bugginess of ML, the post-processing stuff and lack of cheap wide-angle glass, that drove me to the BMPCC.  EOS-M can still be a rock-star. 

More good news is that after figuring out how to use batch mode in Photivo I put some normal focal length, moire filled video, through it using Amaze and 8-pass processing.  It does a great job on the color distortions.  If we can get the EOS-M to shoot in normal focal length without too much moire things will get VERY interesting I dare say :)

The EOS-M still have the same benefits (awesome photo mode, etc.) and drawbacks (ML bugs and post-processing headaches).  The BMPCC hasn't made me want to give up on the EOS-M.  Not by a long shot, as nice as it is.


a1ex

Correction: since there's no ML for EOS M, there can't be any ML bugs ;)

maxotics

Quote from: a1ex on November 23, 2013, 03:11:56 PM
Correction: since there's no ML for EOS M, there can't be any ML bugs ;)

Hi Alex, I'm sure you'd be the last one to argue with me on this.  Most people don't know "official" from non-official ML.  On another forum, a guy posted that the reason Canon doesn't go into this more, RAW for their DSLRs, is that their reputation would be tarnished by all the complaints.  Even if Canon did it well, people would complain that the footage is "horrible and washed out", or "my Class 10 card doesn't work", or "I shot 20 5-minute clips, back-to-back, of my kids play and the camera melted apart" etc :)

In many posts in other threads I did my best to try to organize people around the need to agree on a common set of goals and how to spread out the time to do them.  I'm glad to see you're back, if you are back, but now it looks like 1% has disappeared. 

Anyway, my major point is the EOS-M can be a killer cameras.  Better than the 5D3, because it is so small and versatile.  But if people aren't going to rally behind your issues, which I agree are very serious and important, then what hope is there for the EOS-M  :(


AnthonyEngelken

Super quick question for anyone in the know; emailed Canon, and their reps had no idea. What ISO on the M is effectively the zero gain mark? On the Epic it's 800, but I can't find anything conclusive about Canon APS-C systems and the M. I'd like to start locking the gain at zero, and controlling exposure more with t-stop, lighting, and nd filters.

[edit] It seems the 7d has the best SNR and highest dynamic range at ISO 100, so I'll assume this as my answer because of the shared sensor between the two cameras?

cpreston

Quote from: a1ex on November 23, 2013, 03:11:56 PM
Correction: since there's no ML for EOS M, there can't be any ML bugs ;)

This is unfortunate and depressing.  Magic Lantern is the reason I am still using Canons.  I've used early versions on my 60D and payed on the early release of ML 2.3, which is a wonderful piece of software.  I am now using TL on my 6D and EOSM, both of which don't seem to have a proper ML release.  And, based on experience, I don't trust TL.  I will usually swap in an unmodified card if I know I need to depend on the footage.  I wish I could code as I would be willing to help out on getting ML for both of these cameras.  I would donate more money if I thought it would help.  Even though I've been following these threads I am still not sure what happened here, but I would be extremely grateful if there was support for ML on these cameras. 

maxotics

Quote from: cpreston on November 25, 2013, 06:05:07 AM
Even though I've been following these threads I am still not sure what happened here, but I would be extremely grateful if there was support for ML on these cameras.

The two roots of the problem is too few developers and lack of users organizing themselves to agree on priorities (and too few users doing documentation, etc).  If Alex could clone himself we'd have a stable version for the EOS-M.  From what I can understand, 1% decided he could get more done for the EOS-M if he spend most of his time on his fork.  If 1% could clone himself, it would be stable.  Alex and 1% have had discussions on these issues elsewhere on the forum.  No solution seems imminent.

Everyone, including myself, probably thinks the same thing before trying  to get more done, "do I, can I, commit to the time it will take over the next 6 months to do X?"  Ultimately, everyone seems to kick this can down the street.

I have put in oodles of time developing a C# program to interpolate around focus pixels for the EOS-M.  The other developers were very responsive, but not helpful.  Again, everyone has a list a mile long. 

I got so frustrated with all this I bought a BMPCC.  It's a joy not to deal with all the buggy issues on the EOS-M.  However, it's a great sadness too because, as you know, the EOS-M is incredible and with a stable version of ML, would be the utility camera everyone would want.   I feel I've done my part with the pixel software and shooter's guide.  Jerrykil has been posting builds.  Gary tests everything.  And we have others...so drawing blank of names.  But we only have "one" 1%.  Alex and G3gg0, as much as they would like to help, are focused on other cameras/software.

People have tried to pay them, but I think they reason they're doing all they can.  If they take the money they're just shifting their attention from one person's request to another.  Who knows?






gary2013

TL ver Nov 23, I shot a little bit of 720p24 raw and then I used the ML view function in the File Manager. The data is displayed along with the video, but at the top left it shows the date 25,10,2013 and today is 25,11,2013. I do have the camera set correctly for 25,11,2013??

Gary

jordancolburn

Quote from: a1ex on November 23, 2013, 03:11:56 PM
Correction: since there's no ML for EOS M, there can't be any ML bugs ;)
I've been using the unsupported build for a while now,  and the features and stability (raw video aside) seem to match up pretty well with my ML on other cameras I've used it on (t3i, 60d, etc.) .  Raw video is obviously experimental across all platforms, but it seems to me that the core ML feature set functions pretty well across most cameras.  a1ex's comment (and please correct me if I'm reading too much into this) makes it seem as if there's no real interest in tying it all together into a new official release even though eos m is featured on the ML home page.

Also, RAW is a neat experiment, but don't forget how many great documentaries and shorts have been made using canon's APS-C sensors and built in codecs.  For most almost all applications, dealing with the built in codecs with not  seriously detract from the quality of your final video.  Do not let a lack of usable raw scare you making great videos at an awesome price on this tiny camera with all the video features of the t2i through 7d.

maxotics

Quote from: jordancolburn on November 25, 2013, 09:50:45 PM
Do not let a lack of usable raw scare you making great videos at an awesome price on this tiny camera with all the video features of the t2i through 7d.

We like our RAW self-flaggelation ;)  Anyway, the RAW is very usable.  As you say, you just have to be careful and you're right, the H.264 is as good as any other camera I've tried.

Some REALLY GOOD NEWS EVERYONE, some guy, Joe Jungmann, have no idea who he is, emailed me a vastly improved source to the FocuxPixelFixer.  Before it was slow.  Now it screams!

You can get it here.  Again, only works for 1280x720 for now. 

https://bitbucket.org/maxotics/focuspixelfixer/downloads


gary2013

yes, it is faster. can you add in some extras like creating a new folder using the name of the original raw file and then apply raw2dng or raw2dng_cs2x2_ahdlike_noise to extract the dng files in that new folder?

Gary

jerrykil

Quote from: maxotics on November 25, 2013, 10:36:21 PM
We like our RAW self-flaggelation ;)  Anyway, the RAW is very usable.  As you say, you just have to be careful and you're right, the H.264 is as good as any other camera I've tried.

Some REALLY GOOD NEWS EVERYONE, some guy, Joe Jungmann, have no idea who he is, emailed me a vastly improved source to the FocuxPixelFixer.  Before it was slow.  Now it screams!

You can get it here.  Again, only works for 1280x720 for now. 

https://bitbucket.org/maxotics/focuspixelfixer/downloads
awesome news, maxotics!

jordancolburn

Quote from: maxotics on November 25, 2013, 10:36:21 PM
We like our RAW self-flaggelation ;)  Anyway, the RAW is very usable.  As you say, you just have to be careful and you're right, the H.264 is as good as any other camera I've tried.

Some REALLY GOOD NEWS EVERYONE, some guy, Joe Jungmann, have no idea who he is, emailed me a vastly improved source to the FocuxPixelFixer.  Before it was slow.  Now it screams!

You can get it here.  Again, only works for 1280x720 for now. 

https://bitbucket.org/maxotics/focuspixelfixer/downloads

Don't get me wrong, it's great fun, and I'll still keep trying out raw video every so often as tools improve, but right now, it just isn't workable enough for me yet.  I tried the new pixel fixer, but it crashes halfway through, although oddly enough it seems like the top 1/3 and left 1/3 of the image are free from focus pixels, as if it got partway done before it crashed.  My raw files were 1280, 16x9 aspect, so 1280x434 before the stretch, is this correct?

maxotics

Quote from: jordancolburn on November 26, 2013, 02:18:48 PM
Don't get me wrong, it's great fun, and I'll still keep trying out raw video every so often as tools improve, but right now, it just isn't workable enough for me yet.  I tried the new pixel fixer, but it crashes halfway through, although oddly enough it seems like the top 1/3 and left 1/3 of the image are free from focus pixels, as if it got partway done before it crashed.  My raw files were 1280, 16x9 aspect, so 1280x434 before the stretch, is this correct?

Sorry jordancolburn, the PixelFixer will only work on 1280x720, what you get in crop mode at 16:9 (which will add a 4.5 crop to your focal length).  I plan to make it work for 1280x434 at some point, but don't, because the aliasing/moire in that mode is pretty bad on that camera.   Though I think that can be improved (long story).  For the SD cards, I strongly advise shooting in crop mode.  I get good results with my 10-20mm Sigma EF on adapter.

This is my favorite video with the EOS-M.  It's low light, but I prefer the natural/grainy colors, looks like 16mm film to me, then what I get from H.264 (though I like that better in bright/event situations). 



I agree, RAW isn't for everything.  But when skin tones and shadow detail is important, RAW really delivers.

I should put a warning into the software that prevents it from running non 1280!  Will do.  Thanks for the tip.

jordancolburn

QuoteSorry jordancolburn, the PixelFixer will only work on 1280x720, what you get in crop mode at 16:9 (which will add a 4.5 crop to your focal length).  I plan to make it work for 1280x434 at some point, but don't, because the aliasing/moire in that mode is pretty bad on that camera.   Though I think that can be improved (long story).  For the SD cards, I strongly advise shooting in crop mode.  I get good results with my 10-20mm Sigma EF on adapter.
Oh, so after enabling the 3x crop I will be able to select true 1280x720 without the squeeze in raw_rec?  I just assumed crop mode meant the way it crops toward the center when recording raw video.  I'll try it out soon.


maxotics

Quote from: jordancolburn on November 26, 2013, 04:35:09 PM
Oh, so after enabling the 3x crop I will be able to select true 1280x720 without the squeeze in raw_rec?  I just assumed crop mode meant the way it crops toward the center when recording raw video.  I'll try it out soon.

Don't know if you've read my shooters guide: http://www.magiclantern.fm/forum/index.php?topic=8825.msg82944#msg82944

The reason the vertical is 434 pixels tall is that the sensor has a 4:3 geometry, so it's not tall enough to provide the 720 columns by 1280 rows.

The benefit of crop mode is that is uses a block of CONTIGUOUS pixels in the center of the sensor.  The bad part is the increased effective focal length.

Here is the relevant part from the Shooter's guide:

In full sensor mode, video is generally captured by scanning pixels from a 5,194 by 2,903 (16:9 aspect) of the sensor.  The right way to do it, would be to down-sample all those pixels.  But for electronic/engineering reasons, the EOS-M doesn't read and sample all the pixels, but reads them every other line (The 5D3 does, BTW).  So it skips about 2.2 lines vertically and 7.2 pixels horizontally.  This is complicated by the fact that the sensor pixels each read either a red, green or blue value.  So two lines, if they don't have enough color information (because it's missing from the middle), will end up with a bright red or blue or green.  When the pixels get de-bayered (the full colors calculated) chromatic aberrations occur.  The bottom line is that all consumer cameras, in video mode, suffer from this "line-skipping" problem, to a greater or lesser degree.

In addition to chromatic aberrations, line skipping means that if you have a hard vertical or horizontal line in your image, and it falls between the samples lines, it appears as a jagged artifact.  They form ugly "moire" patterns.

What software is used to de-bayer the image, also has a bearing on the quality.  The distortions from line-skipping show up more in some software and less in others.  I believe Davinci Resolve is the best at dealing with it, but I have no direct experience. 

CROP MODE

In the 19:6 aspect ratio, the EOS-M, can create images from the center 1280 pixels long (wide) by 720 pixels high.  The reason that crop-mode videos do not suffer from moire is it doesn't skip lines, or pixels.  It uses the center-most 1280x720 pixels of the sensor.  The drawback, is that it taking an image from a small part of sensor and enlarging it, so to speak, which creates a zoom, long focal-length; that is, if a lens is 20mm normally, by taking the a center part of the image it becomes 79mm. 

jordancolburn

Thanks for the info, I had read the guide and used crop mode w/ h.264 on the M and on my t3i, but I guess I just misunderstood crop mode in raw to mean the built in crop from using a smaller resolution and not actual crop mode.  Looking forward to trying it out soon.

maxotics

Quote from: jordancolburn on November 26, 2013, 08:22:12 PM
Thanks for the info, I had read the guide and used crop mode w/ h.264 on the M and on my t3i, but I guess I just misunderstood crop mode in raw to mean the built in crop from using a smaller resolution and not actual crop mode.  Looking forward to trying it out soon.

Yes it is confusing.  Crop mode was used in the beginning, with H.264, to boost your effective focal length.  Cheap way to make a 150mm out of 50mm :)  The image is so compressed in H.264 that you don't notice the line-skipping (very much).  In RAW, however, if you blend away all the chromatic aberrations from line skipping you might as well just stick with H.264.   With crop mode, however, you don't get the line skipping, though now the increase in focal lenght works against you when you want wide shots.  Hope that further clarifies. 

Once you shoot some RAW in 1280x720 crop (I suggest a tripod), you're gonna have a revelation!

Escaperoute

I've been using crop mode from the start, with 22mm lens low light video is more than outstanding, i can get some very low light shots with 22mm lens at f/2 but the downside is no IS, but still after your pink dot remover program my life change completely, Now i can produce some really good shorts with my eos-m... Please keep up the good work

jerrykil

Quote from: Escaperoute on November 27, 2013, 08:22:42 PM
I've been using crop mode from the start, with 22mm lens low light video is more than outstanding, i can get some very low light shots with 22mm lens at f/2 but the downside is no IS, but still after your pink dot remover program my life change completely, Now i can produce some really good shorts with my eos-m... Please keep up the good work

i got around to applying a new build of cr2hdr to some old exceedingly noisy photos i took at night and the results are fabulous. thx to those workin on dual_iso. i think its a1ex and marsu. the update have been absolutely great

drchagas

I'm using Build20131012 from max's starter set. I've noticed that when shooting raw video in crop mode, the crop boundary overlay is misaligned.

Has this been fixed in the latest builds? Am I doing something wrong?

EDIT: Solved my problem for now (I probably should upgrade.) Here's my solution, it's not pixel perfect but it's fairly accurate. Haven't tested it at different focal lenghts, I'm using 18mm. Save it to your /ML/cropmks/ folder to use.

https://googledrive.com/host/0B4jFn5c8x4h1djBvbEcxZ0VzUGM/EosmCrop.bmp

EDIT2: Changed the pics to links.

maxotics

Quote from: drchagas on November 28, 2013, 08:09:30 AM
I'm using Build20131012 from max's starter set. I've noticed that when shooting raw video in crop mode, the crop boundary overlay is misaligned.

EDIT: Solved my problem for now (I probably should upgrade.) Here's my solution, it's not pixel perfect but it's fairly accurate. Haven't tested it at different focal lenghts, I'm using 18mm. Save it to your /ML/cropmks/ folder to use.

drchagas, that looks very interesting.  the 11/09/2013 build should be in my latest starter set.  I haven't seen 1% around much lately, and don't believe the build has been changed much in a month, maybe Jerykill or 1% can weigh in here. 

Joe has made even further improvements to the FocusPixelFixer.  I need to incorporate it in some scripts, or an app, to allow one to process a RAW file straight into DNGS, TIFFs or Proress, etc. 

About a post above, yes the 22mm in crop mode gives a great image, but lack of IS, when it's at an effective 90mm focal length, is a downer.  I really wish Canon would start selling the 11-22mm STM lens in the U.S. at a reasonable price! 


feureau

Quote from: maxotics on November 28, 2013, 02:30:18 PM
Joe has made even further improvements to the FocusPixelFixer.  I need to incorporate it in some scripts, or an app, to allow one to process a RAW file straight into DNGS, TIFFs or Proress, etc. 

And .mlv too? :3

jerrykil

Happy thanksgiving, US folk,
maxotics, looking at the commit log, there hasn't been much done as far as rawrec. there was a commit for the intervalometer and MLV things tho

maxotics

I know g3gg0 has been putting tons of time into MLV.  I haven't tried it on the EOS-M. 

1. So is MLV stable on the EOS-M (I only tried it on my 50D)
2. What benefits does/can it give us?