TurboNeRF Releases Version 0.0.11 and 0.0.10

James Perlman released two new updates to TurboNeRF back to back days with 0.0.11 and 0.0.10. See below for the release notes and here for previous versions.

James is also hosting a round of Picture NeRFect later today as part of Edit Fest!

Download

A new version of the TurboNeRF core binary is now available! PyTurboNeRF 0.0.11

A new version of the Blender addon is now available. TurboNeRF-Blender 0.0.11

TurboNeRF 0.0.11

1. Training Raymarcher step size - the lower the value, the higher the resolved NeRF detail will be, but the less rays per step. The higher the value, the lower the detail will be, but more rays can fit in each batch. Under certain circumstances the scene may appear to train more quickly with a larger step size. This control can be found in the Settings section of the TurboNeRF panel in the 3D View

2. Live Preview Raymarcher - similar idea. A larger step size leads to a faster preview, but lower resolution reconstruction. This control is found in the render settings in a new Turbo NeRF panel (it is in the same part of blender where you'd go to change the Render Engine, and only shows up for TurboNeRF. More stuff coming to this area soon).

3. Final Render Raymarcher - located in the same spot as #2. This affects only the raymarching step size in the final render.

TurboNeRF 0.0.10

• There are some optimizations to training that make it faster.

• There is also the capability to render multiple NeRFs, though it's not super usable yet. You can pretty much just duplicate a NeRF for now.

• You can also take your NeRF and rotate/scale/move it around the scene instead of just moving your camera. This is all leading up to the ability to have multiple NeRFs rendered together.

• There is also a special new training control for selecting pixels in datasets with an alpha channel - you can change the Alpha Threshold and the Selection Probability. With both of these controls, you can choose "what percentage of transparent pixels (under a certain threshold) get selected for training." By adjusting these you can often achieve convergence in much fewer steps for certain datasets.

  
  

• A new version of the TurboNeRF core binary is now available! PyTurboNeRF 0.0.10

• A new version of the Blender addon is now available. TurboNeRF-Blender 0.0.10