Recent activity
Subscribe to this feed
Dane Jasper replied on October 22, 2008 16:55 to the question "Thanks for noticing!" in Sonic.net:
Dane Jasper asked a question in Sonic.net on October 22, 2008 16:55:
bitplane replied on October 13, 2008 13:36 to the idea "Recreate a 3D scene by High Definition video camcorder" in Microsoft Live Labs:
I've done this with a video of my town's high street with some success, I stuck a video cam to the driver's side window and recorded it off to MPG format.
I then opened the video in VirtualDubMod and chose "video->frame rate->decimate by->8" and "file->export image sequence" and created a stream of windows bitmaps.
Then, because PhotoSynth doesn't support bitmaps I had to install ImageMagik to convert the images, which gives me the command line tool "convert.exe"
To do this I opened a dos prompt and used "dir /b *.bmp > textfile.txt" to create a list of files, pasted the list into a columns B and C of a spreadsheet (OpenOffice Calc because I don't own Microsoft Excel, but you could also use Google Docs), replaced ".bmp" with ".jpg" in column C, and put the text "convert -quality 95" into column A.
I then copied the 3 columns and pasted them into notepad, saved as "converter.bat" and double clicked it.
The results were reasonably good with 640x480 and 900+ images, going round corners broke the synth because I couldn't aim at a reference point with the camera stuck to my window! Photosynth seemed to think the road was bent in a fish-eye fashion, but I'm assuming more data points would fix this
However, now I have the technique I'll try using it to create the point cloud for a fully synthed area before adding high quality photos over the top. It would be really cool if I could hide the low quality images from the viewer, and just use the video for mapping.
Keith replied on October 10, 2008 16:32 to the question "when will skitch for windows be available?" in plasq:
tahrey replied on October 05, 2008 14:38 to the question "Does the size of the photo (megapixles) matter" in Microsoft Live Labs:
Jon, Nate - yes, that one. Some of them do seem to load up OK if you leave them a while but others (e.g. as in my attachment) either never fully load, or are taking so damn long about it that no-one, even myself, will hang around long enough to allow them to do so. Also the initial load-up seems to take forever, enough so that I was motivated to comment on it in my synth description. Is it just overly complex and causing issues either on the server or the flash client? Or held on a server that's having load / connection issues that haven't yet been picked up on ..... or suffered untrapped processing / uploading difficulties that have caused this effect. (Possibly something to do with that unreadable info-page also winding up synthing to completely the wrong wall for reasons that can't easily be explained other than an all-out glitch - one that I think has also affected at least one of my other large synths in a couple of odd ways)
As before, I'm not hugely bothered by it as I largely achieved what I wanted to when embarking on the experiment, but if they are supposed to be loading up quickly into their full, 2 megapixel, medium-high ISO, slightly blurred glory, then it's definitely something that needs looking into!
Thanks...
(EDIT: ah... it's auto-reduced that ... if there's any way you can zoom into it, however, the effect should be plain... in fact i think it's slightly noticable even here - if only because of the rather stark jpg compression it's also added to the image. I measured it in Paint as displaying at 680x512 pixels or thereabouts (within the maximised 1024x768 window), and it's still distinctly blocky, so it may even be 320x240 rather than 400x300!)
PS that was after leaving it sit for a couple of minutes, when a couple other pictures in the set on the "walk" towards it appeared in much better definition after only ~10 seconds (still seeming a bit slow, but did get there) ... oddly my impression of the dark, translucent surrounding pics on some of the locations was that they were clearer than the one in the centre!
In fact, it still hadn't resolved after I'd saved that image and shut paint again.
A comment on the question "Does the size of the photo (megapixles) matter" in Microsoft Live Labs:
Jonathan, I'm fairly certain it's this one. The first photo doesn't seem to do too badly, but the none of the shots of the courtyard or the photos on the wall seem to resolve fully. – Nathanael Lawrence, on October 02, 2008 23:41
A comment on the question "Does the size of the photo (megapixles) matter" in Microsoft Live Labs:
can you add the link to the synth you feel shows the images drastically below their uploaded resolution? I'd like to take a look.
thanks – Jonathan, on October 02, 2008 22:51
tahrey replied on October 02, 2008 22:46 to the question "Does the size of the photo (megapixles) matter" in Microsoft Live Labs:
Sooooo.... in that case, if it can deal with the pics at their native resolution and display them nicely, and synths at about 1.5mpx (i'm assuming 1600x900, 1536x1024 or 1440x1080 depending on aspect?) --- how come one I did with a few hundred 2mpx images looks for all the world when displayed that they've been reduced to 400x300 or less? I originally didn't shoot at anywhere near my camera's top resolution (8mpx) but chose a setting I thought would give a good, clear result that could be zoomed in a little if desired, without using up megatons of storage space, processing power or upload time.
It's pretty obvious that a major resolution reduction has taken place, as it's of some artful documentary photos on display at my workplace (using a glass-sided corridor as a naturally well lit gallery, and including walks up and down said corridor with the not-entirely-successful intent that you can act as a virtual visitor strolling up and down) and the poor definition is immediately evident on viewing any of the photos themselves... some of them actually being difficult to make out properly now... and most particularly the notice giving information about the exhibition. It's on a standard piece of printer paper at about 14 or 16 point text, which I know is perfectly legible at 2mpx (1600x1200) as I used to use my first, 2mpx-maximum digital camera quite successfully as a low-tech document scanner for material printed on the same size paper but with smaller point sizes. The page as displayed via photosynth is utterly unreadable, it's just a blocky blurry mess, you can't even make out the title that's a couple points larger and in bold.
I was accepting of it maybe having to reduce the picture fidelity in order to cope with larger numbers of pictures or synths that were all-round more complex, and it being more an advert for the expo rather than a reproduction of it (its not like I have the copyright on the thing anyway - but it is being shown for free at a public access site, so all I'm really doing is teleporting people to that location via the internet) until I read the company blurb above saying that it should give a perfect reproduction of the original good-detail pictures. Now I'm wondering what's going on, and at which point the failure lies - within the app, or within the propaganda?
It's not just that they're needing a long time to load up on a slow connection or something, and therefore only getting so far through the uninterpretable -> vaguely recognisable -> blocky but obvious -> reasonable -> full resolution stages that are plain to see most of the time. It gets up to a somewhat blocky state and then "hangs" there without getting any clearer, even if left for some minutes, and the same occurs when you move on to a new picture - the only reasonable assumption is that's actually the resolution they've been stored at on the server. The point cloud is pretty awesome and easily recognisable as the place I have to walk up and down every working day, but the photo quality itself leaves something to be desired.
Any clues?
I'm not bothering to do my future big-synths in anything higher than this setting until we figure out what's up here :) and in fact I'll probably downconvert them to XGA to save more space. (Using VGA as source is a loser's game, I've found out that much)
(not that I would probably have done them in higher rez anyway, though my last one I took all at 5mpx before rendering down to XGA just to get a bit more sharpness in each pixel and reduce the effect of CCD noise)
A comment on the question "Does the size of the photo (megapixles) matter" in Microsoft Live Labs:
JackH - does your mouse have a wheel? – Dane Jasper, on October 01, 2008 22:56
Johan Benesch's reply to "Merge point groups to increase 'synthiness'" was just promoted to the most useful! Dane Jasper and 2 other people think it's one of the best replies.
i recently also tried to synth a building.
when i redid the synth with many more pics to better cover the corners
the synthiness dropped ...
since you have pretty good idea which "orphan synths" that have overlap it would be great to be able to give that input to the software.
could be as simple as drag-and-drop one pic in one "orphan synth" to a good overlap with a pic in another "orphan synth".
Dane Jasper replied on October 01, 2008 22:54 to the question "Timeframe for update allowing 'Add Photos/Edit' options?" in Microsoft Live Labs:
David,
One thought. The challenge is two areas that don't meet because there were not quite enough transitional images between them. Your suggestion to take more photos and re-synth is fine, IF the scene is one that can be revisited.
However, often it cannot. For example, a vacation, or something temporary like an event. Or, something which has changed.
The ability to give the synther some hints to help it connect points it can't quite work out sure would be nice.
-Dane
Dane Jasper replied on October 01, 2008 22:45 to the question "Same synth input with different results?" in Microsoft Live Labs:
Dane Jasper marked one of David Gedye's replies in Microsoft Live Labs as useful. David Gedye replied to the question "Same synth input with different results?".
David Gedye replied on October 01, 2008 21:54 to the question "Same synth input with different results?" in Microsoft Live Labs:
Photosynth can give a slightly different result if you give the images to it in a different order. The reason for this is that the reconstruction phase of the algorithm is an optimization problem that can never be perfectly solved, and final results depend on certain choices of an initial pair of images to start with. If you give the images in a different order, then it starts "growing" a consistent 3D solution from a different seed, and the results will be slightly different.
Does that help?
Dane Jasper marked one of Nathanael Lawrence's replies in Microsoft Live Labs as useful. Nathanael Lawrence replied to the idea "Linking synths?".
A comment on the question "Same synth input with different results?" in Microsoft Live Labs:
It does seem logical. I'll be curious to hear what Live Labs has to say about it. – Nathanael Lawrence, on October 01, 2008 15:09
Dane Jasper replied on October 01, 2008 15:07 to the question "Same synth input with different results?" in Microsoft Live Labs:
Nathanael Lawrence replied on October 01, 2008 14:53 to the question "Same synth input with different results?" in Microsoft Live Labs:
-
Nathanael Lawrence started following the question "Same synth input with different results?" in Microsoft Live Labs.
-
Nathanael Lawrence started following the idea "Linking synths?" in Microsoft Live Labs.
| next » « previous |
Loading Profile...







