Skip to content

Latest commit

 

History

History
92 lines (59 loc) · 1.98 KB

DATA_PREP.md

File metadata and controls

92 lines (59 loc) · 1.98 KB

Dataset Preparation

Waymo

To work with Waymo Open Dataset, make sure to register on their official website and install Google Cloud CLI

1. Authenticate gsutil to your account

  • It will generate a verification code for you to copy and paste in
gcloud auth login
gcloud auth list

2. Configure gsutil to your account credentials

gsutil config

3. Install the dataset

Original Training Dataset (~1 TB)

make download-waymo-full

Small Subset of Training Dataset (~5.8 GB)

make download-waymo-mini

CO3D

To work with a lightweight subset of the CO3D dataset (Common Objects in 3D), follow the steps below. These instructions are adapted from the official CO3D GitHub repository.

1. Create the dataset directory

Create the directory in the current project folder

mkdir -p data/co3d

2. Clone the CO3D repository

Clone the CO3D codebase outside of your current project folder:

git clone git@github.com:facebookresearch/co3d.git
cd co3d/

3. Install dependencies

Install the required Python packages:

pip install visdom tqdm requests h5py

Then install the CO3D package itself:

pip install -e .
  • Note: Make sure to install these packages in a separate environment

4. Download the small subset of the dataset

Use the CO3D download script with the --single_sequence_subset flag to fetch a compact subset suitable for the many-view, single-sequence task:

python ./co3d/download_dataset.py \
  --download_folder DOWNLOAD_FOLDER \
  --single_sequence_subset

Example (downloading into this repo’s data/co3d folder):

python ./co3d/download_dataset.py \
  --download_folder ../Open-Rig3R/data/co3d/ \
  --single_sequence_subset

This subset requires ~8.9 GB, significantly smaller than the full dataset (~5.5 TB).