Due to the omnipresence of Neural Radiance Fields (NeRFs) the interest towards editable
implicit 3D representations has surged over the last years. However editing implicit or hybrid
representations as used for NeRFs is difficult due to the entanglement of appearance and
geometry encoded in the model parameters. Despite these challenges recent research has
shown first promising steps towards photorealistic and non-photorealistic appearance edits.
The main open issues of related work include limited interactivity a lack of support for local …

In Fig. 1, we show the network architecture for LAENeRF. As discussed in our main material,
our module is NeRF-like: The sizes of the hash grid and linear layers as well as the
spherical harmonics encoding are used similarly in iNGP [11]. However, instead of two
subsequent MLPs predicting density and color, we split our model and evaluate the weight
and the offset network concurrently.