SE(3) Equivariant Convolution and Transformer in Ray Space

PyTorch implementation of paper "SE(3) Equivariant Convolution and Transformer in Ray Space"

Environment

This work follows the architecture of the work: IBRNet: Learning Multi-View Image-Based Rendering and we use the same environment. To create an anaconda environment:

conda env create -f environment.yml
conda activate lightfield

Datasets

We use the same datasets as IBRNet, please follow the instruction to download the datasets.

Running

example of multiple GPUs:
python -m torch.distributed.launch --nproc_per_node=8 train.py --config configs/train.txt

example of a single GPU:
python train.py --config configs/train.txt