Skip to content

Latest commit

 

History

History
465 lines (465 loc) · 12.8 KB

data_train.md

File metadata and controls

465 lines (465 loc) · 12.8 KB

Data

You need to follow directory structure of the data as below.

${ROOT} 
|-- data
|   |-- base_data
|   |-- ARCTIC
|   |   |-- data
|   |   |-- preprocessed_data
|   |   |-- splits
|   |   |-- dataset.py
|   |-- Decaf
|   |   |-- data
|   |   |-- preprocessed_data
|   |   |-- splits
|   |   |-- dataset.py
|   |-- DexYCB
|   |   |-- data
|   |   |-- preprocessed_data
|   |   |-- splits
|   |   |-- toolkit
|   |   |-- dataset.py
|   |-- H2O
|   |   |-- data
|   |   |-- preprocessed_data
|   |   |-- splits
|   |   |-- dataset.py
|   |-- H2O3D
|   |   |-- data
|   |   |-- preprocessed_data
|   |   |-- YCB_object_models
|   |   |-- dataset.py
|   |-- Hi4D
|   |   |-- data
|   |   |-- preprocessed_data
|   |   |-- splits
|   |   |-- dataset.py
|   |-- HIC
|   |   |-- data
|   |   |-- preprocessed_data
|   |   |-- splits
|   |   |-- dataset.py
|   |-- HO3D
|   |   |-- annotations
|   |   |-- data
|   |   |-- preprocessed_data
|   |   |-- dataset.py
|   |-- HOI4D
|   |   |-- data
|   |   |-- preprocessed_data
|   |   |-- splits
|   |   |-- dataset.py
|   |-- InterHand26M
|   |   |-- annotations
|   |   |-- images
|   |   |-- preprocessed_data
|   |   |-- splits
|   |   |-- dataset.py
|   |-- MOW
|   |   |-- data
|   |   |-- preprocessed_data
|   |   |-- splits
|   |   |-- dataset.py
|   |-- ObMan
|   |   |-- annotations
|   |   |-- data
|   |   |-- object_models
|   |   |-- preprocessed_data
|   |   |-- splits
|   |   |-- dataset.py
|   |-- PROX
|   |   |-- data
|   |   |-- preprocessed_data
|   |   |-- splits
|   |   |-- dataset.py
|   |-- RICH
|   |   |-- data
|   |   |-- preprocessed_data
|   |   |-- splits
|   |   |-- dataset.py

base_data

bash scripts/download_train_base_data.sh

preprocessed_data

  • Download preprocessed_data from HuggingFace by running:
bash scripts/download_train_preprocessed_data.sh

ARCTIC dataset

${ROOT} 
|-- data
|   |-- ARCTIC
|   |   |-- data
|   |   |   |-- cropped_images
|   |   |   |-- images
|   |   |   |-- meta
|   |   |   |-- splits
|   |   |-- preprocessed_data
|   |   |   |-- train
|   |   |   |   |-- annot_data
|   |   |   |   |-- contact_data
|   |   |-- splits
|   |   |   |-- train.json
|   |   |-- dataset.py

Decaf dataset

${ROOT} 
|-- data
|   |-- Decaf
|   |   |-- data
|   |   |   |-- train
|   |   |   |   |-- cameras
|   |   |   |   |-- ...
|   |   |   |   |-- videos
|   |   |   |-- test
|   |   |   |   |-- cameras
|   |   |   |   |-- ...
|   |   |   |   |-- videos
|   |   |-- preprocessed_data
|   |   |   |-- train
|   |   |   |   |-- contact_data
|   |   |   |   |-- db_bb.npz
|   |   |   |   |-- db_contact.npz
|   |   |-- splits
|   |   |   |-- train.json
|   |   |-- dataset.py
  • Download data by running:
bash scripts/download_official_decaf.sh
  • preprocessed_data, splits are already downloaded from HuggingFace.

DexYCB dataset

${ROOT} 
|-- data
|   |-- DexYCB
|   |   |-- data
|   |   |   |-- 20200709-subject-01
|   |   |   |-- ...
|   |   |   |-- 20201022-subject-10
|   |   |   |-- models
|   |   |-- preprocessed_data
|   |   |   |-- train
|   |   |   |   |-- annot_data
|   |   |   |   |-- contact_data
|   |   |   |-- test
|   |   |   |   |-- annot_data
|   |   |   |   |-- contact_data
|   |   |-- splits
|   |   |   |-- train_s0_29k.json
|   |   |   |-- test_s0_5k.json
|   |   |-- toolkit
|   |   |-- dataset.py
  • Download data except data/models and splits by running:
bash scripts/download_official_dexycb.sh
  • Download data/models by running:
bash scripts/download_official_ycb_models.sh
  • preprocessed_data, splits are already downloaded from HuggingFace.

H2O dataset

${ROOT} 
|-- data
|   |-- H2O
|   |   |-- data
|   |   |   |-- label_split
|   |   |   |-- object
|   |   |   |-- subject1_ego
|   |   |   |-- subject2_ego
|   |   |   |-- subject3_ego
|   |   |   |-- subject4_ego
|   |   |-- preprocessed_data
|   |   |   |-- train
|   |   |   |   |-- annot_data
|   |   |   |   |-- contact_data
|   |   |-- splits
|   |   |   |-- train.json
|   |   |-- dataset.py
  • Download label_split.zip, object.zip, subject1_ego_v1_1.tar.gz, subject2_ego_v1_1.tar.gz, subject3_ego_v1_1.tar.gz, subject4_ego_v1_1.tar.gz from official website and place them in data.
  • preprocessed_data, splits are already downloaded from HuggingFace.

H2O3D dataset

${ROOT} 
|-- data
|   |-- H2O3D
|   |   |-- data
|   |   |   |-- train
|   |   |   |-- evaluation
|   |   |   |-- train.txt
|   |   |   |-- evaluation.txt
|   |   |-- preprocessed_data
|   |   |   |-- train
|   |   |   |   |-- contact_data
|   |   |-- YCB_object_models
|   |   |-- dataset.py
  • Download and extract H2O-3D v1 from official website and place it in data.
  • If you have not already run this script for DexYCB, download YCB_object_models by running:
bash scripts/download_official_ycb_models.sh
  • preprocessed_data is already downloaded from HuggingFace.

Hi4D dataset

${ROOT} 
|-- data
|   |-- Hi4D
|   |   |-- data
|   |   |   |-- pair00
|   |   |   |-- ...
|   |   |   |-- pair37
|   |   |-- preprocessed_data
|   |   |   |-- train
|   |   |   |   |-- annot_data
|   |   |   |   |-- contact_data
|   |   |   |-- test
|   |   |   |   |-- annot_data
|   |   |   |   |-- contact_data
|   |   |-- splits
|   |   |   |-- train.json
|   |   |   |-- test.json
|   |   |   |-- train_pid.json
|   |   |   |-- test_pid.json
|   |   |-- dataset.py

bash scripts/download_official_hi4d.sh
bash scripts/extract_official_hi4d.sh
  • preprocessed_data, splits are already downloaded from HuggingFace.

HIC dataset

${ROOT} 
|-- data
|   |-- HIC
|   |   |-- data
|   |   |   |-- 01
|   |   |   |-- ...
|   |   |   |-- 12
|   |   |   |-- IJCV16___Results_MANO___parms_for___joints21
|   |   |   |-- HIC.json
|   |   |-- preprocessed_data
|   |   |   |-- train
|   |   |   |   |-- annot_data
|   |   |   |   |-- contact_data
|   |   |   |-- test
|   |   |   |   |-- annot_data
|   |   |   |   |-- contact_data
|   |   |-- splits
|   |   |   |-- train.json
|   |   |   |-- test.json
|   |   |-- dataset.py
  • Download data by running:
bash scripts/download_official_hic.sh
  • preprocessed_data, splits are already downloaded from HuggingFace.

HO3D dataset

${ROOT} 
|-- data
|   |-- HO3D
|   |   |-- annotations
|   |   |   |-- HO3D_train_data.json
|   |   |   |-- HO3D_evaluation_data.json
|   |   |-- data
|   |   |   |-- train
|   |   |   |-- evaluation
|   |   |-- preprocessed_data
|   |   |   |-- train
|   |   |   |   |-- annot_data
|   |   |   |   |-- contact_data
|   |   |-- dataset.py
  • Download and extract HO-3D v2 from official website and place it in data.
  • Download annotation files from HandOccNet and place it in annotations.
  • preprocessed_data is already downloaded from HuggingFace.

HOI4D dataset

${ROOT} 
|-- data
|   |-- HOI4D
|   |   |-- data
|   |   |   |-- datalist
|   |   |   |-- HOI4D_annotations
|   |   |   |-- HOI4D_CAD_models
|   |   |   |-- HOI4D_cameras
|   |   |   |-- HOI4D_color
|   |   |   |-- HOI4D_Handpose
|   |   |-- preprocessed_data
|   |   |   |-- train
|   |   |   |   |-- annot_data
|   |   |   |   |-- contact_data
|   |   |-- splits
|   |   |   |-- train.json
|   |   |-- dataset.py
  • Download and extract HOI4D_annotations.zip, HOI4D_CAD_Model_for_release.zip, camera_params.zip, HOI4D_release.zip, HOI4D_Hand_pose.zip from official website after download request and place them in data.
  • Download datalist by running:
bash scripts/download_official_hoi4d.sh
  • preprocessed_data, splits are already downloaded from HuggingFace.

InterHand2.6M dataset

${ROOT} 
|-- data
|   |-- InterHand26M
|   |   |-- annotations
|   |   |   |-- train
|   |   |   |   |-- InterHand2.6M_train_camera.json
|   |   |   |   |-- ...
|   |   |   |   |-- InterHand2.6M_train_MANO_NeuralAnnot.json
|   |   |   |   |-- InterHand2.6M_train_data.pkl
|   |   |   |-- test
|   |   |   |   |-- InterHand2.6M_test_camera.json
|   |   |   |   |-- ...
|   |   |   |   |-- InterHand2.6M_test_MANO_NeuralAnnot.json
|   |   |   |-- val
|   |   |   |   |-- InterHand2.6M_val_camera.json
|   |   |   |   |-- ...
|   |   |   |   |-- InterHand2.6M_val_MANO_NeuralAnnot.json
|   |   |   |-- skeleton.txt
|   |   |   |-- subject.txt
|   |   |-- images
|   |   |   |-- train
|   |   |   |-- test
|   |   |   |-- val
|   |   |-- preprocessed_data
|   |   |   |-- train
|   |   |   |   |-- annot_data
|   |   |   |   |-- contact_data
|   |   |-- splits
|   |   |   |-- train.json
|   |   |-- dataset.py
  • Download annotations, images by running:
bash scripts/download_official_interhand26m.sh
  • preprocessed_data, splits are already downloaded from HuggingFace.

MOW dataset

${ROOT} 
|-- data
|   |-- MOW
|   |   |-- data
|   |   |   |-- images
|   |   |   |-- masks
|   |   |   |-- models
|   |   |   |-- poses.json
|   |   |   |-- watertight_models
|   |   |-- preprocessed_data
|   |   |   |-- train
|   |   |   |   |-- contact_data
|   |   |   |-- test
|   |   |   |   |-- contact_data
|   |   |-- splits
|   |   |   |-- train.json
|   |   |   |-- test.json
|   |   |-- dataset.py
  • Download images, models, poses.json by running:
bash scripts/download_official_mow.sh
  • masks, watertight_models, preprocessed_data, splits are already downloaded from HuggingFace.

ObMan dataset

${ROOT} 
|-- data
|   |-- ObMan
|   |   |-- annotations
|   |   |   |-- obman_train.json
|   |   |   |-- obman_test.json
|   |   |-- data
|   |   |   |-- train
|   |   |   |-- test
|   |   |   |-- val
|   |   |-- object_models
|   |   |   |-- watertight_meshes
|   |   |-- preprocessed_data
|   |   |   |-- train
|   |   |   |   |-- annot_data
|   |   |   |   |-- contact_data
|   |   |   |-- test
|   |   |   |   |-- annot_data
|   |   |   |   |-- contact_data
|   |   |-- splits
|   |   |   |-- train_87k.json
|   |   |   |-- test_6k.json
|   |   |-- dataset.py
  • Download and extract obman.zip from official website after download request and place it in data.
  • Download annotations and splits by running:
bash scripts/download_gsdf_obman.sh
  • object_models, preprocessed_data, splits are already downloaded from HuggingFace.

PROX dataset

${ROOT} 
|-- data
|   |-- PROX
|   |   |-- data
|   |   |   |-- quantitative
|   |   |   |   |-- body_segments
|   |   |   |   |-- ...
|   |   |   |   |-- sdf
|   |   |   |   |-- vicon2scene.json
|   |   |-- preprocessed_data
|   |   |   |-- train
|   |   |   |   |-- annot_data
|   |   |   |   |-- contact_data
|   |   |-- splits
|   |   |   |-- train.json
|   |   |-- dataset.py
  • Download data by running:
bash scripts/download_official_prox.sh
  • preprocessed_data, splits are already downloaded from HuggingFace.

RICH dataset

${ROOT} 
|-- data
|   |-- RICH
|   |   |-- data
|   |   |   |-- hsc
|   |   |   |-- images_jpg_subset
|   |   |   |-- multicam2world
|   |   |   |-- scan_calibration
|   |   |-- preprocessed_data
|   |   |   |-- train
|   |   |   |   |-- annot_data
|   |   |   |   |-- contact_data
|   |   |   |-- test
|   |   |   |   |-- annot_data
|   |   |   |   |-- contact_data
|   |   |-- splits
|   |   |   |-- train.json
|   |   |   |-- test.json
|   |   |-- dataset.py
  • Download data by running:
bash scripts/download_official_rich.sh
  • preprocessed_data, splits are already downloaded from HuggingFace.