Can’t seem to get something to synth right? Curious about what lens is best? Ask your fellow Photosynthers here.
Is there anyway to manually synth photographs? The images are in the correct order and they overlap to make the synth work, but the auto synth is not working. Is there a way for me to manually syth them up?
Do you have an example synth that you think didn't match up well? There is currently no way to manually help the synther match photos.
I had the same experience with my first synth. If I could control the order of the pictures I am sure the auto synth would make a better job of it. So how about being able to arrange the placing and order in the grid view?
Admittedly, my pictures were more suited to a montage than a true synth, but the results could definitely have been better.
We certainly haven't spent time trying to make a great expeience for groups of photos (like Bigessay's) that aren't very synthy.
A manual editing program, where people get to hang their own photos in whatever 3D locations they like is possible, but we suspect that only a tiny fraction of the people who use Photosynth automatically, would be prepared to do manaul 3D positioning. That's why it's not very high on our wish list right now.
Are we wrong?
@Bigessay: I must admit that even I had trouble seeing right away how some of those photos overlapped. Unfortunately Photosynth needs to take things in smaller steps than a human brain. Some of those shots, though they contained the same objects, viewed them from radically different angles. Pages 6 and 7 of the Photosynth Photography Guide ( http://photosynth.net/howtosynth.aspx ) cover what angles should result in good synthesis.
@David: I think assuming that all Photosynth users form the same demographic is not necessarily right. Or, put another way, I think that when people experience their photos successfully arranged spatially as Photosynth does, the frustration that they didn't all match is, in many cases, equal to the satisfaction or joy that is experienced when they have.
It does feel like a bit of a tease to have such an amazing way to view photos and then have the last few paraded in front of you without a way to place them correctly within their peers.
Out of curiosity what would you set as lower barrier for including an image in synth at? Would just removing singletons be enough? Would you only want to include the main 3D group? Or do you envision something even more granular? Remember this would need to be set on the synther, before you've had a chance to view the synth.
I'd love to have a toggle where I could say, "Only present the main 3D group. Drop the rest." Most of the time I'm building a synth that's what I have in mind. (I'll post this to the feature request forum...)
Tony, I'm not sure I understand that first sentence.
While I agree that only uploading/displaying 'Synth 0' would make many people happy, it is fundamentally different than allowing users to manually 'fix' the photos that didn't synth in automatically.
I realise that what I'm about to reference seems to rely on a successful synth as the first step, but what I was hoping for was something along the lines of the interface seen in Drew Steedly's work with Rick Szeliski and company here: http://research.microsoft.com/en-us/um/redmond/groups/ivm/PlanarStereo/ .
Although I am interested in generating models as they detail, I would really be far more interested in manually aligning the photos that didn't automatically synth using that same interface used to notate common surface planes among photos. Ideally the synther could use that plane data as a basis for the positioning of the photos that can't be deviated from very far and then finesse it using feature matching.
I'd love to be able to 'help' the synth, perhaps give it hints on images that I know are a related and contain common elements.
I've been using images from a library archive (some very old images) and sometimes it works whilst other times it doesn't and I wish i could give it just that extra nudge :)
Btw, although some of these archival images don't synth all that well (the long dead photographers have clearly not even bothered to read the guide) when it does work it really is quite a startling sensation.
p.s. the video on the guide page isn't showing.
Yeah, adding the possibility to pick where to add the given not synthed photo would probably help the program as that would point where photo should fit, thus lowering the automatic thresholds.
P.S. and the video i sstill unavailable.
P.S.2. Great software nonetheless :)
Lapies, both versions of the video have been back up for some time now. I just tested both of them again they're playing fine on my computer.
Perhaps you need a Windows Media Player plugin for your browser (or Flip4Mac if you're visiting the site from OSX)?
I've taken a series of photo of a "cactus garden". I've taken them to use with software that "compose" them on a 360deg panorama so probably there is enought overlapping to generate a good synth but success rate is only 47%... don't really understand why. It could be that program have problem with images where there aren't too much building or easily indeitificable points ?
Here you can see the synth:
It would be nice if we could help synth by just clicking matching points in multiple pictures. So that we would like suggest where is the match, and he has to gues the 3D meaning of it. Creating synths would go in two phases. 1st Synth creates an automatic version and than user helps him refine it by clicking similar points between different groups.
I agree with brongracz. Other 2D panoramic photostichers (e.g. Canon Photostitch) provide this. It would be excellent if Photosynth did as well.
Makan, I'm a big proponent of getting manual intervention in Photosynth and have been for a while (2008 August 30: http://getsatisfaction.com/livelabs/topics/timeframe_for_update_allowing_add_photos_edit_options#reply_587240 ), but it honestly doesn't sound like the Photosynth team is interested in providing this to us, as much as I wish that they were.
I don't know if you've already seen the following discussion already or not, but I put together a brief explanation on why shooting a synth is different than shooting a panorama for 2D stitching software here that references Photosynth guidance material, if you're interested: http://photosynth.net/discussion.aspx?cat=ceca0f30-0f7c-4468-811a-32b623ea8563&dis=363ec255-07ab-4675-8dfc-4eed62a69f45
There are solutions out there that allow you to manually choose keypoints in multiple images, but many are fully manual with none of Photosynth's automatic identification and matching.
OpenPhotoVR is one such manual solution. ( http://openphotovr.org/ )
Photofly from Autodesk is more of a hybrid of automatic and manual, like I wish Photosynth was.
Here's an introductory video: http://www.youtube.com/watch?v=ttaQ7WyVbEE
And here's the link to their site: http://labs.autodesk.com/technologies/photofly/