Homework 1
PBRT Setup & Installation
The PBRT installation was trivial and easy on Mac OS X (10.4.3 / Build 8F46). I had to manually build and install libtiff to compile the exr tools (exrtotiff, tifftoexr etc). This was also relatively painless. I had to slightly tweak the typical configure based buid cycle (configure, make & make install) to build libtiff on OS X but I found directions online for this and didn't run into any problems. I also downloaded and built OpenEXR, only to later realize that Preview displays exr files and so I could just ues the Finder to view them.
Process & Observations
I had read about the Camera space coordinates in the book and so didn't have any difficulty setting up and positioning lights. However, I did find configuration 1 tedius to reproduce since I had to reduce the cone angle to keep the backdrop from showing (I actually don't understand why the sample image is setup this way since the assignment handout made no mention of this). Configuration 2 was straightforward to setup. For configuration 3 it took me a few tries to figure out what parameters to use for my fill light so that the image would not be overexposed. Configuration 4 was reasonably tricky because I had to also figure out which lights were spot lights and which were area lights. It was obvious that the fill light ought to be an area light while the accent light should be a spot light. However, the other two lights could either be spot lights or area lights or one of them could be a spot light and the other an area light. I tried just using spot lights and got acceptable results so I just stuck with them.
In general, like everybody else I produced lower resolution images with a small number of samples per pixel and per area light to keep my test-render cycles fast. For the head model I used a single sub-division count while testing and rendered with 2 for my final images. When comparing against the samples in the handout I found it best to first convert my exr images to tiff since I didn't know what the equivalent exr images should look like and there was definitely a visible difference between the tiff and exr files.
Results & some Extras
I was having fun with pbrt so I tried a few more materials and settings. Here are low-resolution images that I created. I used blue plastic to come up with the "blue man" look; clay to produce a sculpted look and a skin texture for the model and sky texture for the backdrop to add some realism to the scene. Without subsurface scattering, hair and the oily sheen of skin, this latter image still seemed very fake. I also tried creating the T2 liquid metal look and think the glass material worked pretty well for this. Finally, I thought I'd try to use a photon integrator and produced some images with direct, indirect and caustic photons. The last 3 images show these effects. The first image in this set is produced normally, the second uses indirect & direct photons and the last is built using caustic photons only.
