Extreme point cloud, 1000 images used, (bridge model, iasi)

this is a much better reconstruction than the earlier version i looked at. can you describe in more detail what you're not satisfied with? if you're looking for more points on the bridge beams itself you probably won't get them. the feature detection doesn't focus on edges like our eye, it tends to prefer textures. the beams don't have much texture so taking more photos probably won't help.

I like your goal of "Only 100% synthy" on your profile.

Is this a friend helping you take pictures of the bridge: http://photosynth.net/view.aspx?cid=0b3ef704-3cf4-4872-9ec4-559f0c829552&i=0:0:632&z=27.139859823062746&g=0&p=-2.81349e-041:-2.71373e-071&m=true

@ madeeds: Well, you guessed, thoose shots in a blueys tint and low louminosity are made by a friend of mine which unfortunatelly hasn't got a decent camera phone. Mine's were taken on another phone :P se w810. I'm quite happy with the quality. Would have been nicer to take 3 MP instead of 2 MP photos.

@ K: there are 3 things that i'd like to corect:

1) Bridge top, pyramid apex, has a lot of noise | 20-30 photos pointed at the top of the bridge with the horizon line at the bottom of the frame should do the job ...

2) Bridge deck is almoust invisible | 100-150 photos walking twice across the bridge in two paralel rows from one end to another and back, camera orientation slightly downwards should help increase point density also maybe more well defined safety rails ( it hwlped a lot the 2 rows of shots pointed at the safety rails)

3) Background buildings are too blury | 100-200 photos walking on the banks ~200 meters in each direction camera pointing on a

perpendicular axis with the axis of the river. THis should help make a nice background

4) And lastly removing the 400 pics that don't help too much.

I'll have to wait a little for the right wather cause now it's snowy and photosynt will not make anny sense of the photos :P

It's going to be hard to ever get a point cloud of the bridge deck, unfortunately. A point requires a feature, and a feature requires some sort of texture. Though the deck is textured, the texture is so repetative that the synther has trouble distinguishing one brick from another, so it has a hard time positioning it in 3-space. You could do some experimentation to see what works here. I don't think that walking in two different directions will help as much as walking the same direction two or three times in the same direction on paths parallel to eachother.

Back and fourth would'nt really help for deck plane but would help for the guard rails. Also the pavement isn't 100% repetitive. There are a lot of features for synth to grab on to like eroded bricks, the gap in the middle, metal joints for seasonal cycle of dilatation/contraction and the guard rails themselfs in relation with the background buildings.

I'm still waiting for better weather. Winter has returned around :P

Btw, there are a few false features in the point cloud. If you look carefull enough you will spot a big cloud of points, white and red, floating just alongside the bridge above waterline. I know that they are from the second set of photos I tooked with my camera. After a lot of examination a saw that it has similar features to the neighbouring leg. I think a few more shots to that leg will solve that problem.

Let's say that's the east side of the bridge. If you look in the south side you'll see snaped to the bridge central axis after the building facade below groundlevel, a plane of beige points. Also after a closer inspection I saw yellow points on the edges. I think that those might be part of the deck plane: beige>pavement, yellow>guard rails. Puzzleing me is how did they got there?

Also weard stuff underground bebeath water surface and in the north side...

Use the nav keys to see theese problems more easely.

Nice synth! After seeing the bridge, it makes me wonder why I haven't seen more designed that way. It's a really cool way to support it. I guess it would only work for fairly small gaps.