Gabriel Walsh's dashboard

David,

I love the concept...recreating a 3d past from old photos.

Having looked at the photos you are trying to use (and with the caveat that I have no technical backup for this opinion and have no affiliation with the Photosynth or LIve Labs team) my gut feeling is that they are going to be unlikely to produce a good comprehensive synth either unedited or edited as you suggest. They just feel like there aren't enough consistent shots in your collection yet from the perspective of similar shapes, angles, colors, textures, etc.

That said though, the idea intrigued me enough that I screen captured 9 of the photos from your collection that seemed the likeliest to me to synth well in a perfect world and then:
1) normalized their colors to somewhat of an "old photo" style
2) cropped them a little to get rid of the worst offending differences
3) tried to remove artifacts that might throw everything off
4) re-uploaded to this location: Test-Hong Kong

.

Unfortunately, as you can see here, only 2 of the resulting photos ended up properly synthed. Now, the fact that I didn't have the originals, or the fact that I only spent a couple minutes normalizing the images, or the fact that I cropped the images and color adjusted them could certainly affect the end result (according to what the Photosynth team has said elsewhere), but in this case I think it is more likely that the raw materials just aren't good enough to accomplish the goal.

I hope I am wrong though and that you figure out a way to do it (or perhaps are able to find a number of other/better samples from closer angles that synth better.

Good Luck

Sherwood,

Okay, the funniest thing about your question is not at all what most people will think upon first read...no, the funniest part of your question is that someone named Sherwood is asking a Forest question! That is priceless!

That said, and David’s very relevant Caveat notwithstanding for the moment, your inquiry made me wonder just what your hypothetical might actually require to accomplish in terms of time, photos, distance, etc., so here’s my attempt at some basic math.

- An Acre is about 44,000 sq. ft. That means the Hundred Acre Wood comes in at about 4,400,000 sq. ft.

- Assuming the Wood was a perfect square (which I don’t think it was, but the math should stay pretty much the same anyway), that makes it about 2,100 ft on a side.

- Based on your “every 50 feet” specification, that means there would be about 42 photo opportunities per pass, with about 42 passes, for a total of 1,764 photo opps.

- Based on your “5 photos” per stop specification, that makes about 8.820 photos per full coverage pass of the Wood.

- At this point, you mention making two criss-cross passes of the wood, but I have to ask you a question...if you are taking pictures in all directions, what good would a second pass really do? If you are looking for more overlap, why not just take 8 photos per stop and save yourself the extra pass? If you are looking to criss cross in opposite directions, won’t you still end up in essentially the same photo spots? If not, then why not either a) shoot NE, SE, SW, NW at each stop also, or b) stop every 25 feet? For argument’s sake though, let’s assume the two passes you hypothesize at the outset: that makes for a total of 17,640 photos.

- At about 5MB each (10-12 Megapixels, I assume?), that’s about 88 GB of storage.

- As for the distance/time, assuming 42 – 2100 ft passes, you’re looking at about 17 miles per pass, for a total of about 34 miles of walking.

- Assuming an average walking pace of about 3 MPH, you’re looking at just over 11 hours of walking

- We still need to add in photo time, so even if you’re really quick, I think we need to bank on at least 5 seconds per shot (this is a very low estimate, but no reason to exaggerate here, right!? ;) . This equates to about 24 hours straight taking pictures.

- Total time: approximately 35 hours.

Which leaves me with a few parting thoughts:

- If a photographer falls in the forest, does he make a sound?

- Maybe you should think about synthing the Honey Tree first? Or Tigger? (Stripes are good for synthing, I Imagine!)

Good luck on your quest!

Gabriel

David, my two cents on this one are: cent 1) "fully resolve" is definitely better than don't resolve at all and move on too quickly, but it may still leave a little to be desired.
cent 2) As much as your uncluttered interface / settings concept is great, it is still important to allow the creator of the content to dictate the user experience. Here's what I mean: flash has few if any settings the end user can realistically "adjust", but the creator of an swf can add in an infinite amount of control: automatic slideshow speeds, speed controls, hotspts to allow advance, etc, etc. So, what I think is critical, ESPECIALLY in the embed mode of Photosynth, is that the creator is allowed to dictate things like (among others) the play speed of the slideshow. This will immediately make the tool a fair amount more powerful to different types of content creators and therefore likely increase adoption significantly. – Gabriel Walsh, on September 02, 2008 02:16

Nathanael et al, Now that I have tested it, I am realizing that your point about the camera position info not being available through the embed code is exactly the problem I am having. I have tried to use the hybrid of the two methods you mentioned (i.e. using the http link provided in the email option after the embed.aspx from the option to embed), but I have not had any success. Have you heard anything from the "Photosynth team members" you mentioned above? This seems pretty critical to me (although I'm sure we all think that about our ideas!) and I don't see the all-important "The company is considering this idea" notation, so I'm reaching out again in hopes that there either 1) already is a way to do this that I just don't realize, or 2) will be a way to do this soon? Anything you or your team members can come up with to help would be great. Thanks again. – Gabriel Walsh, on September 02, 2008 01:37

In most cases I have seen with this, (especially the ones over the last few days) the problem seems more likely to be a referesh issue of the client side application/dialog box than an actual issue with publishing. I have seen the "stuck at publishing" issue many times (probably 5-10 or so times in 20-30 test synths I have done). In all of the recent ones, the synth showed up on the site (i.e. I could go to "My Synths" and it would be there completed and visible) within 10-30 minutes of when it said "publishing" on the application/dialog, but often the dialog lags behind hours or indefinitely forcing me to close it. Perhaps the issue is now related (or at least sometimes related) to the app just not refreshing the status of the site appropriately rather than the publishing not happening?

Yeah...fair enough, and I think I agree for the most part, but that wasn't the core of the thought, just one potential use. The core is the ability to alter a photo for creative puurposes without confusing the synth tool. – Gabriel Walsh, on August 30, 2008 13:58

Not directly, Breezest, but there are a couple Powerpoint plugins that might accomplish what you need. LiveWeb http://skp.mvps.org/liveweb.htm and T'dog http://www.tdogsoftware.com/ have tools that can put any web page on a Powerpoint slide. I tried one here in LiveWeb and it worked fine (although it was the entire Photosynth page, not just the iFrame embed to it). Having said that, I have always been a fan during PowerPoint of having a browser open separately and using <alt> <tab> to toggle between presentation and other content. This avoids the problem of having a bad webpage or other problem crash my primary presentation, and I think it generally looks seemless enough that it shouldn't cause much of an issue.

I guess ithis might be obvious also, but you could certainly just hyperlink text or even an image screenshot of your synth in Powerpoint and have that launch IE separately. I'm guessing from your post though that you want it truly embedded for some reason. In that case, I think the two tools I mention above are probably your best bet.</tab></alt>

Great Idea Gavin...This is exactly the thought I was planning to post until I saw your comment. I understand the need for the scaffolding but also agree that it would be great to focus greater attention on some images than others. Another way to think of it would be two buckets : one for pics used in the display view and one for those that aren't. Theoretically you could allow either the creator, the end user, or both to move photos between the two buckets, right?

Nathanael, Thanks for the great feedback. Even though I think it would be a good bonus to be able to set the actual first image as mentioned, I think you make a great argument here that the camera postion embedding covers most if not all of the hypothetical situations I am picturing. – Gabriel Walsh, on August 27, 2008 14:16