5,336
edits
Instant Neural Graphics Primitives with a Multiresolution Hash Encoding: Difference between revisions
Instant Neural Graphics Primitives with a Multiresolution Hash Encoding (view source)
Revision as of 17:51, 31 January 2022
, 31 January 2022→Method
(→Method) |
|||
| Line 16: | Line 16: | ||
# Pass <math>\mathbf{y}</math> through your small feature decoding neural network. | # Pass <math>\mathbf{y}</math> through your small feature decoding neural network. | ||
===Levels=== | |||
For resolutions in <math>[N_{min}, N_{max}]</math>. | |||
* Resolution at level l: <math>N_{l} = \lfloor N_{min} * b^{l} \rfloor</math> | |||
* <math>b = \exp \left( \frac{\log N_{max} - \log N_{min} }{L-1} \right)</math> | |||
===Hashing Function=== | |||
* Hash collisions are ignored. | * Hash collisions are ignored. | ||
* The hash function used is: <math>h(\mathbf{x}) = \left( \oplus_{i=1}^{d} x_i \pi_i \right) \mod T</math> | * The hash function used is: <math>h(\mathbf{x}) = \left( \oplus_{i=1}^{d} x_i \pi_i \right) \mod T</math> | ||