Corinne Alini HW 2

I have used 0 late days on this assignment

Question 1.3: Ray Sampling

Xy Grid

Ray Bundle Question 1.3

Question 1.4: Point Sampling

Points Sampling Question 1.4

Question 1.5: Volume renderingg

Question 2.2: Loss and Training

Box center: (0.25, 0.25, -0.00) Box side lengths: (2.00, 1.50, 1.50)

Question 2.3:Visualization

Question 3: NeRF

Question 4.1:

We are able to render more details when we add view dependence. In the images viewed below you can see that the wheel of the lego is rendered better with view dependence than without.
Regular (no view dependence):
Question 4.1

View Dependence:
Question 4.1

This will not generalize as well. We are training the network to seek out finer details, so it will have issues generalizing with new examples as the finer details will be different.

Question 4.3:

With the higher resolution, it takes way longer to run and way more memory. In fact, gpu ran out of memory when I tried to run the higher resolution with n_pts_per_ray equaling 128. This image demonstrates the higher resolution image with n_pts_per_ray being 64. You can see that the image is blurrier as we do not have as many points to compare it to. Question 4.3

I also played around with shrinking the hidden layers. You can see that it is far blurrier than the example above which makes sense as it is learning fewer details. It does nto learn the color of the lego as well. You can see that it isnt as vibrant and doesnt learn the red color. It loses the finer details of the features.
Question 4.3

The tradeoff is accuracy vs memory usage/time to train.