Can’t seem to get something to synth right? Curious about what lens is best? Ask your fellow Photosynthers here.
# of pictures doesn't matter?
Synthiness score is the "connectedness" of the photos. If they are broken into two equal size clumps, you'll get 50%. If you have one clump of 18 photos and 2 that matched together you might get 91%. When viewing the synth, click on the grid view button (nine little squares) to see which pictures matched together and which didn't.
In short, yes, number of pictures is independent of synthiness score.
Still I'm not to sure how this works. How do you go about to create a good synth? I have done panormic shots with up to 5 shots in succession following one plane but my Synth attempts have been mediocre at best. Is there a sequence you have to load the shots or will the synth happen anyway like happens for panoramic photos when Live Photo Gallery stitches them together? Also are there any plans to have this application as a downloadable stand-alone application like the afore-mentioned Live Photo Gallery?
It's a shame that your synth isn't public or linked here in this topic, otherwise I'd give it a look and let you know what specific problem you're dealing with.
The order that you load images in can affect the outcome of the synth a very small amount when there are loads of images but if you're putting relatively few in a synth then you shouldn't have too much of a problem or notice much of a difference.
As far as making a 'good synth', well there are lots of different goals. Do you only care that all the photos match and it tells a good story or has beautifully composed|framed photos and that there aren't too many photos to get lost in the middle of? Or, are you the sort of person who only cares about getting a good point cloud or model of what you're synthing and the photos are not of much interest to you?
Both of the approaches (+ all the mixes of the two) above have their time and place but will lead you to shoot in very different styles.
My advice for a beginner who wants to see something cool successfully pop back out of Photosynth is to pick an object that you can walk all the way around (nothing too shiny) and that you can back away from a little without anything blocking your view and just take photos in a circle all pointing at the one object.
Like this: http://photosynth.net/view.aspx?cid=4eca7430-533a-44af-b991-1534310ea71b
(Note that you don't need to take anywhere near the number of photos that I did. It's all about the movement of the camera: how much your photos overlap each other and don't change angles too much too fast.)
Practise this method on a few single objects and see how they go. Once you get the hang of it, you can expand it to something like this:
This synth, although 2x as big, didn't take that long to shoot + is no more complex than taking photos circling single flowers + looking from one to another.
The number of photos that you can synth together is only limited by *2* things.
1: How synthy is the thing that you're photographing? (More rough or textured will get you more points for your pointcloud sooner which is good. If you're trying to use heaps of photos (I'm talking about close to 1000 or more) this will limit how many you can put together because more image features per photo means run out of memory sooner, trying to figure them out.)
2: How powerful is your computer? (More than 2GB of RAM is good and you should have plenty of free space on your hard drive. A faster processor will help you get the job done faster, so that if your computer munches away on a huge number of photos for a while and then runs out of memory at the end of making the synth and cancels the whole thing - you will have wasted less time waiting for it to finish. Again, this should only be a concern once you start exceeding the 300 photo limit suggested by the Photosynth team.
As far as having a Photosynth type program that you can just use on your own computer... don't hold your breath. Sorry :)
Firstly, Microsoft is very interested in being able to link together synths of outdoor places and the indoors of public places like museums, art galleries, restaurants, etc. This can only happen if everyone's synths keep being published online where they (Microsoft) can someday hook them together.
Secondly, in a synth each photo is actually a little larger than the original photo (about 1.5 times as large) because it is actually now the original resolution + a half as large version, another a quarter of the original size, etc. All sizes are divided into tiles and zipped for fast loading off of a hard drive or over the internet without worrying about loading any full image (you only load the tiles you need to look at any part of each photo). This would quickly eat up your hard drive space if you were saving all your synths on your own computer
One last note: for the synths that I linked to above, use the fullstop or period key to walk through the photos in the order that I took them. The spacebar or next icon will take you through Photosynth's idea of what a good order is, but using the [.] key will show you exactly how I approached the shooting.
If you or a friend has an iPhone or iPod Touch you should download the iSynth app as there is an option in it to show all the camera positions onscreen. This is an incredibly useful tool if you find a great synth and want to see how the author approached their subject to get the result that they did.
Here's the link: http://www.cs.brown.edu/people/gpascale/iSynth/
This may have been covered, but here's my $0.02 on getting high synthy numbers:
Photosynth works off of parallax. In an ideal world every feature of the scene will have been photographed from at least three vantage points. This gives three separate sets of angles to establish the location of that point in 3D space and establish the location of the camera for each image.
If you want to do a panoramic synth, do three panoramas of the same scene from three locations. Make sure every part of each panorama overlaps with the other panoramas. The wider the angle between your three locations and the subject, the tighter the fit to each point's location in space, and the better the solution for the camera location.
As has already been pointed out, you need texture so that key points can be established between images. Featureless expanses of color, objects that move, WATER, these things do not contribute at all to a synth.
Once you have a set of pictures that conforms to this, go through it and chuck blurries, pictures that don't really tie in well to neighbor pictures, etc. Then go back and double-check to make sure you have at least three views of each feature in the scene from three different angles.
Regardless of whether you use 30 images, 300, or 1000, it's the interactions between the images that's going to determine how synthy the set is rather than the actual number of images. If you have a 30 image synth and every single image can be tied into the 3D frame of reference with good connectedness to neighbor photos, it'll be 100% synthy.