Skip to content

Latest commit

 

History

History
25 lines (22 loc) · 1.28 KB

DATASET.md

File metadata and controls

25 lines (22 loc) · 1.28 KB

Dataset Preparation

CO3D v2

  1. Please follow the official instruction to download the dataset.
  2. We use the dataset provider in Implicitron for data loading. To speed up the loading, we cache the loaded meta data. Please run
cd scripts
python prepare_co3d.py --co3d_path [path to CO3D data]

to generate the cache. The cached data take ~4.1GB of space.

Hypersim

  1. Please follow the official instruction to download the dataset.
  2. We preprocess the Hypersim data for faster loading:
cd scripts
python prepare_hypersim.py --hypersim_path [path to Hypersim data]

The resulting data take ~19GB of space.

  1. We evaluate our method using the ground-truth meshes, which are available for purchase here. We borrow the dataloader from Omnidata for loading the meshes. To speed up dataloading, we preprocess the data by
python prepare_hypersim_mesh.py --omnidata_starter_dataset_path [path to omnidata starter dataset] --hypersim_path [path to Hypersim data]

The resulting data take ~5.5GB of space.