Conquistadores Field Trip Report

On Friday November 2, a few of us went down to Little Holes Canyon near the San Rafael River to check out the Conquistadores as a field work test site for our upcoming modeling project. After a little "navigational ambiguity" on the rim of the canyon, we found our way over to the Conquistadores and determined that this is a near perfect test site for us.

We want some rocks which are visually complex, easy to get to but require traveling on foot and afford 360 degree access. We got all of that plus 360 degree access from above (on the canyon rim) and below (on the canyon floor).

The geometry is a little intimidating, but that's what makes it research. We thought it would be important to photograph some of the funkier shapes so that potential reviewers would know that the shapes were real and not an artifact of the modeling process.

Natural Phenomena in... "Surfs Up!"

I watched Sony's Surf's Up! the other day on the small screen. The use of natural phenomena was refreshingly ambitious most of the time and comically over-ambitious at others. Overall, it improved my impression of Sony's Animation group. A group of Sony Animation VPs and others came out to BYU for a visit a few weeks ago. They had good feedback for my student Cory Rheimschussel's cloud effects in "Kites" (mostly "that's hard, good luck") and the concept art for "Cloudy With a Chance of Meatballs", some of which was done by a BYU animation intern, looked great.

Back to "Surf's Up!"... it was good to see directors use weather, wind and time-of-day in a story. Scenes on the beach front, generally, included trees moving in the ocean breeze. A few scenes were actually set in weather other than sunny and clear. The rain was nicely done as was the sea fog. To my recollection, this is the first film in which trees moved in the wind and the weather was part of the set. Nice job.

Early on, there was a scene in which Cody rode a whale out of Shiverpool. Unfortunately the water-to-foam transition on the whale's wake was done poorly. Laughably poorly. I felt like I was watching a concept shot in dailies. I was surprised that it made the final cut. The transition seemed to be implemented as "when the curvature passes horizontal, switch immediately to foam!"

Possible Field Work Sites

We are looking for a field work test site somewhere in the desert. We need a place with the following characteristics:

1. Within 2.5 hours of Provo, Utah.
2. Includes a cool looking rock formation. Need to walk around the entire rock formation.
3. Low elevation. We want year-round access.
4. Is between 1 and 2 miles from the closest road. This is negotiable though.

Got any ideas?

We are thinking of using the Conquistadors near Little Holes Canyon near the Grand Canyon of the San Rafael Swell, see picture above. The problem with the Conquistadors is that they are next to a 50 foot cliff with loose rock. Safe enough, but safer would be better.

We are also thinking of using the first Pinnacle (following Steve Allen's guidebook) in Pinnacle Canyon off the Tidwell Draw North Road.

Reflectance from Photographs

Here's the problem: we've got an object of known geometry in a known location at a known time. The problem is to get the color and reflectance from a collection of those photographs. That's a large part of the intellectual merit of this proposal we are working on. The photograph shows such an object in my driveway on a Friday afternoon. The cylinder is 12 inches high and has a diameter of 6 inches. We painted it with matte and semi-gloss spray paint. The painted portions will allow us to calibrate our results with surfaces of known color and reflectance.

Virtual Parks++ ?

We are building up steam to launch a project which involves solving parts of the plenoptic function for natural landmarks. The current state of the art for field capture of interactive scene data seems to be capturing cylindrical panoramos using rotating heads on tripods: http://www.virtualparks.org/

Imagine doing the same thing, but supporting non-deforming geometry, novel viewpoints and relighting. Now imagine putting all that technology in a package with free software and sub $10,000 capitol equipment cost and giving it to the kind of people who go record this stuff for fun in their free time.

One concern is that the experience becomes to "Disneyland". That is, the resulting immersive environment is so compelling and complete that the visitor wonders "why bother with all the walking? I've already seen it here on my TV." I read that the other way. I think "now that I've got a glipse of the place, I am highly motivated to protect it and to get out there and experience it for real."

We need to think more carefully about this issue.

SIGGRAPH 2007 paper and link follow up.

We need to follow up on a few SIGGRAPH 2007 papers. and ideas
These are:

• Image clip art. The low-cost approach to modeling lighting in a single photograph is interesting. Also, take a look at the LabelMe database.
• That paper about clipping using columns of pixels that "aren't that interesting" for use in a CS 312 project.
  
• All of the image-based rendering papers
  
• Sphereon technologies
• HD viewer for gigapixel images from MSFT
  
• Polhemus
  
• Voodoo camera tracker
• Syth-eyes camera tracker (maybe)
  

Clouds: Highlights from Mt. Timpanogos (take 2)

We have a long-running data collection project going to record images of Mt. Timpanogos, Utah, using a digital camera connected to a computer. So far, we've got about 5,000 images paired with time of day and weather information . The goal is to learn to relight terrain using this information.

The images posted here are just a few I picked out because they look interesting. Based on these images it appears that getting clouds right is more important than we'd originally thought. The color and texture of the clouds adds a lot to the mood in each of the images.

Clouds: Highlights from several thousand pictures of Mt. Timpanogos

 
 
 
 

We have a long-running data collection project going to record images of Mt. Timpanogos, Utah, using a digital camera connected to a computer. So far, we've got about 5,000 images paired with time of day and weather information . The goal is to learn to relight terrain using this information.

The images posted here are just a few I picked out because they look interesting. Based on these images it appears that getting clouds right is more important than we'd originally thought. The color and texture of the clouds adds a lot to the mood in each of the images.

There's an image with the Alexander's Print Shop balloon in it somewhere in the collection. I found it one afternoon but haven't seen it since. When we find it again, I'll post it.

Goblin, version 5

These two images show one of our goblins superimposed on a photograph from Goblin Valley. The geometry of the goblin was generated from a block of voxels using a spheroidal weathering algorithm. Original photograph in the background used courtesy of Leping Zha, www.lepingzha.com. (Algorithm: Matthew Beardall, Texture and Animation: Cory Rheimschussel).

Goblin, version 4

 

Our goblins have come a long way since January. We use the same basic erosion simulation, but we've got better back-end processing. The back-end processing smooths the surface, otherwise we get lego-like shapes (as seen below). In this version, we are using marching cubes to extract the 0.5 isosurface (0.5 measures the degree of decimation of the volume). We do some more smoothing in Maya and then apply two textures. One texture for the top and one for the bottom. Cory Rheimschussel did all Maya magic and these two textures. Darius Ouderkirk implemented marching cubes.

Natural Phenomena Workshop 2007

The 2007 Eurographics Workshop on Natural Phenomena will be held on September 4 2007 in Prague as part of Eurographics 2007. I was walking around SIGGRAPH last summer thinking to myself "there should really be a workshop on natural phenomena, I think I'll put one together" Then I walked by the Eurographics booth and saw the proceedings from the 2005 workshop for sale by their booth. I immediately bought the proceedings and thought to myself "that was a lot easier than organizing a workshop"

Details: EGWNP 2007

Controlling a Canon PowerShot Camera using PSRemote under Vista

We have a long-running project in the lab in which we are capturing images of mountainous terrain and recording the time and weather in which the image was captured. To date, we have a collection of about 1,500 images paired with time and weather information. We use a Canon PowerShot A620 to collect the images and we use PSRemote by Breeze Systems to script the shots.

On a whim, I decided to install Vista on the machine that takes the pictures. That was a bad idea in the end because Canon has no plans to support Vista with the drivers needed to run PSRemote.

So I installed Virtual PC from Microsoft, put a clean copy of XP on the virual PC and plugged in my camera. Turns out that Virtual PC doesn't support USB devices.

So I grabbed a copy of Virtual Box and repeated the process. Virtual box does support USB connections, but I needed to add Canon's USB explicitly to the USB pass-through filter. That code is 04a9 (courtesy of www.linux-usb.org).

After downloading and installing everything on the Virtual Box, we were back in business and here's an image that was captured last night.

Cubic Goblin

 

The latest and greatest goblin is extremely symmetric and has a nice cube shaped top. The significance of this image is that it shows the results of differential erosion with a hard layer on top and a soft layer on bottom.

We are working on two main issues now. First, getting the results to look less symmetric. Second, exporting the shape to Maya in a relatively smooth fashion for final rendering.

(Credits: Pretty much everyone in the Sandstone Terrain Feature Capstone class for Winter 2007)

Goblin Skin Textures

 

A test render of a Goblin highlighting a skin surface texture by Cory Reimschussel. The small surface features are bump mapped and the colors are intended to match the color of a real Goblin. The geometry is the same geometry generated by Darius Ouderkirk using Matthew Beardall's erosion algorithm. The geometry has two problems. First, there's a slew of infinitely small fins between some faces and it looks like it is made from Lego brand stackable building blocks. The texture isn't too bad and we are working on the geometry as well as the erosion algorithm.

Field Trip Report

The CS Department has posted a report on our recent field trip to Goblin Valley.

I was pleasantly surprised at the value of the field trip. I originally planned it because it might be useful but it would definitely be fun. It turned out to be fun (as expected) but also was more useful than many of the capstone class discussions we'd had prior to the field trip.

Unfortunately, the BYU Rental Fleet Safari vans have a low ceiling which meant I had to dip my head slightly for 6 hours while I drove back and forth between Goblin Valley. This left me with an excruciatingly painful headache after I got home. We'll have to solve that problem next time by renting an SUV from the fleet.

First Image: Erosion with Concavities

This image is our first fully rendered terrain generated using our model of surface erosion that admits concave surfaces. It obviously needs some serious smoothing, but it looks more like a Goblin than our last one. (Algorithm: Matthew Beardall, translation to and mock-up in Bryce by Darius Ouderkirk)

This work was done as part of our BYU CS Capstone class on sandstone erosion modeling.

First Synthetic Goblin

Matthew Beardall developed an erosion simulation algorithm which created the following hoodoo-like shape. If you look at it long enough, you can perhaps convince yourself that the algorithm is on the way to generating Goblin Valley Hoodoos (link goes to an image by Eve Andersson).

Average Sunset on Mt. Timpanogos

This image is the average image created by compositing 43 images of Mt. Timpanogos taken at 99.6% of the lit day over the course of the last 2 months. As such, it represents the average sunset over that time period. The average noon-day picture (taken at 50.0% of the lit day) is bit more murky.

Introducing the brand new Computer Generated Natural Phenomena Lab at BYU

My new lab is the "Computer Generated Natural Phenomena" Lab. I am working with a grad student (co-adivising with Parris Egbert) and two undergrads. We even made a lab webpage to increase our visibility.

Natural Phenomena isn't a hugely hot research area in graphics right now. There's usually a paper in two in SIGGRAPH and there's a Natural Phenomena workshop held with Eurographics each year. I believe that while CGNP is hugely important in visual storytelling, the fundamental algorithms and assumptions haven't changed much in 10 years. Computing technology has come along way since 1996 and it's time to rethink and re-invent CGNP technology. I intend to position my students at the forefront of a second wave of CGNP research.

Contact me if you'd like know more about our work including how you might participate.