2009 Gigapan of Finals

Competition Championship Event
1

This is a 1 Gigapixel image I shot at finals in ATL.

This image was shot as the Einstein final was still setting up.

The first of 3 panos at this location. Others from a slightly better location to be posted later.

I used the Beata version of the Gigapan imager to shoot a 9 X 30 array of 270 images.

Click on the link for an interactive zoomable image that allows faces at the other end of the dome to be recognized.

http://www.gigapan.org/alpha/gigapans/21481-778x200.jpg

See the notes at the above link for camera and exposure details.

I used the free Gigapan stitcher software to assemble into a single image and post it.

I posted this image quick with no cleanup. I will clean it up and repost later. .

I will have 2 more to post once they are rendered. These are later in the finals program with most people seated and also includes more of the upper balcony and VIP’s

Click on the snapshots. It is easy to generate your own snapshots but you may have to get a free account on Gigapan to do be able so. Getting an account is quick and easy. It would be neat to see teams generate snapshots of themselves.

Note that the gigapan stitcher software is free so anyone with a point and shoot camera and a tripod can generate these large images. The Gigapan pan/tilt actuator just simplifies it and makes it much faster to shoot the images.

Note that Austin from Team 2837 (“Total Chaos” on gigapan.org) borrowed the Gigapan imager and shot several good images in ATL as well. See them here:

http://www.gigapan.org/viewProfile.php?userid=17571

EDIT: I have found out that total that Total Chaos (Austin) has a CD thread on his images here:

========================================

Also, Dave Lavery shot some gigapan images in 2008. See them here.

http://share.gigapan.org/viewProfile.php?userid=2946

EDIT: I just found out that Dave was setting up his Gigapan on a ladder at Einstein. I can’t see him in my images but I hope he got an image of my gigapan setup on the balcony.

============================================
GigaPan was developed by Carnegie Mellon University in collaboration with NASA Ames Intelligent Robotics Group, with support from Google.

The idea is to get an inexpensive gigapixel+ imaging system into the hands of millions of people so that they will post on google earth and gigapan.org.

See details at www.gigapan.org

The fully functional and free stitcher software is here:

http://gc.cs.cmu.edu/stitcher/

Regards
Frank Neuperger
Team 39

2 Likes
2

Awesome I love how much I can zoom in! I even found myself :smiley:

1 Like
3

Wow, this is a fantastic picture. It is amazing how far I can zoom in. This picture was taken the perfect moment when I was enjoying some cookies n’ cream Dippin Dots. Thank you so much for this!

1 Like
4

Hey! There’s my team right up in the front of the picture!
I seem to have been really tired when you took this…

Could you share what kind of camera and lens you used? The images are really high res and very nice quality on the zooms, so I’m a bit curious as to what you were using, since my camera hits 480 with the crop factor on my telephoto lens and I could still only barely see the people on the other end of the arena.

5

The camera was an inexpensive superzoom by Panasonic
LUMIX FZ18 ~$250
8 M pixel
set to 18x optical zoom which is 35mm equivalent of 504 mm.
The key is that I took 270 of the 8 M pixel pictures in a 9 row by 30 column array and used the Gigapan stitcher to merge them all.

You loose ~50% pixels in each image because of 30% overlap in each axis.

The merge of the 270 images took 3 hrs on vista 64bit quad cpu (Q8300) WITH 8G 1366 MHz DDR3 memory. The process was set at elevated priority. All 4 cores were buzzing. Have had some of these take 6+ hours to merge.

6

Slightly related topic - does anyone know if anyone has made a photosynth (www.photosynth.net) from inside the Dome?

7

awesome stuff…

im waiting to see dave’s gigapan as well… he was set up on a ladder against the curtain on einstein… i cant wait…

8

Very nice quality pan. I see myself at the other side of the arena, under the overhang where it’s pretty dark, and the image is still pretty clear.

It really took 3 hours to stitch? With a quad and 8 GB ram it should be pretty smooth. What are you running for clocks on it, and on what motherboard? (Sorry for thread derail, but im curious :stuck_out_tongue: )

At any rate, real neat gigapan

9

This is awesome. I was able to zoom in on our team, and find our two students sitting in the awards section!

Incredible.

10

Okay so I found myself in the picture of the pits but not this one. And I was in the awards section. O well. And this is really really nice.:smiley:

11

That deserves my worship. The photo is amazing on so many levels! The entire Blue Cheese team is on there (luckily)!

12

Awesome. Simply awesome. Literally.

13

Here is a quick stitching of one of the Gigapans that I shot from the back of Einstein during the competition finals.


(click on the image to link to the full Gigapan)](http://share.gigapan.org/viewGigapan.php?id=21597)

There is a lot of work still to be done with this image to pull out some of the blurred people and replace them with the clean sections of overlap areas, and balance out the colors. But this is a good first indication of what the final image will look like.

The image has small sections of significant blurring where people were moving rapidly while the sub-images were being acquired. Due to the uneven light coverage inside the Georgia Dome, I was shooting with a 1/8-second shutter speed (Canon G-10 at f4.5, full optical zoom, no digital zoom). This was slow enough to catch considerable movement in each frame. I shot the panorama four times, and have enough duplicate frames to clean up the blurred areas with a little bit of work.

-dave

.

14

Looks great Dave! I love it! Way better than mine, especially in terms of image consistency, and a great camera angle. I’ve already started snapping away taking snapshots of people I recognize that go on ChiefDelphi. :stuck_out_tongue: (Hope you don’t mind)

And Frank, I’m glad to see you got this up right away, I was just about to email you when I saw this thread. It looks great! And it makes me wish I would’ve taken full advantage of my camera’s optical zoom in mine. Oh well, next time! :stuck_out_tongue:

  • Austin
15

Austin,

Yes, for max zoomability in the in the pano, I generally go to the max optical on the camera. There are caveats however. For still subjects, zoom to max, but for crowds that are moving like this one, if you zoom too tight, you can end up with a ton of rework if you want to minimize motion artifacts of missing legs heads etc. The frequency of occurrence of these artifacts goes up as you zoom tighter simply because there are more frames (stitch boundaries) for a given size pano. Notice the right hand side of my image in the first posted pano as the subjects were closer. The stitching really chopped some of those people up because they were large in comparison to the frame and they were not sitting still.

Mechanically the Beta gigapan is so so. Ok for a light point and shoot. Because of the crush of trying to be competitive in ATL, I did not have the time to better integrate the camera with the Gigapan to ensure that the camera c of g was always on one side of the tilt axis for all angles of tilt. As such I got vertical smearing on some images. The shutter servo induced flex due to use of that small plastic spacer instead of the big aluminum spacer.

Also, I did not really fully test the delays associated with my auto focus to see if the gigapan was actuating the shutter before the autofocus settled. . To get great panos, you have to characterize all of the idiosyncrasies of your camera in the context of the Gigapan mechanism. .

I am going to add a bearing support on the STBD side of the Gigapan and remake the camera and shutter servo mounting plate out of 1/8 wall plate or extrusion to get it really rigid. This should also help the tilt axis “stiction” get a bit more deterministic.

After having to build that swerve steered drive for 09 , this will be a snap!!

Having a remote control (even if wired) to pause it for people to walk past would be another great addition. I hate having to touch it in case I push too hard an make the stepper slip. Not sure how to do this. Need to check if the code is open source and if there is some I/O left on the board.

16

James,

q8300 quad core at 2.5 GHz. 8 Gbytes 1366 MHz DDR3 memory.

It is an iBuypower machine. Forgot the mother board. I think the bus is 1600 MHz capable. Too bad the stitcher does not use the 64 core NVIDIA graphics card.

I am looking into setting up a Ramdrive to do the stitching as I think I am bottle-necked at the I/O to the disk. It never gets above 35% average on CPU even when I elevate task priority.

Frank

17

Cool technology. The 3D support structure of the inside of the dome would be very interesting to try with Photosynth. Too bad there is no way to export the 3-D model. Perhaps I will get some photos with parallax next year… if we make it to the Championship.

Note that the Gigapan image sets made from the stationary camera position are not likely to be too useful for Photosynth as the information content of the Photsynth pixel clouds are most likely enriched by the parallax from changing the camera position. I wonder if (and if how) the recent Mars exploration images were processed to extract 3-D models of the terrain and features.

Frank
Team 39

18

Great shot of Team 39

http://www.gigapan.org/viewGigapan.php?id=21598&snapshot_id=65908

http://www.gigapan.org/alpha/snapshots/65908-300x200.jpg

1 Like