jonbarron
jonbarron
Log(scale) shouldn't be thought of as a regularizer, it's the log of the partition function of a probability distribution. Basically, this is not a "design decision", like L2 or L1...
This is a good idea if 1) you believe in maximum likelihood estimation (https://en.wikipedia.org/wiki/Maximum_likelihood_estimation) and 2) if you want to maximize the likelihood of the observed data you're training on.
If the alpha value stays large throughout optimization, it sounds like your data doesn't have very many outliers, in which case you'll probably get optimal performance by just allowing alpha...
There are unit tests that need to be updated as well.
I patched your PR and ran nosetests from the root directory, and it causes the testImageLossfun* tests to fail. Those tests need to also be rewritten with the same NHWC->...
I'm not entirely sure what you're describing, so if you have a colab that reproduces this issue feel free to link it. I would strongly advise against overfitting to a...
Those datasets haven't been released yet, yeah.
My guess would be that your camera poses are incorrect, or are in a different format from what the code expects.
Ignore that warning about normals not being rendered, that is to be expected.
Yep those look like good cameras! Weird.