junukcha / text2hoi

Text2HOI: Text-guided 3D Motion Generation for Hand-Object Interaction

Text2HOI

Official code of Text2HOI: Text-guided 3D Motion Generation for Hand-Object Interaction in CVPR 2024
Arxiv paper / Project page

Data

H2O.

GRAB.

ARCTIC.

MANO.

obj.pkl

data
├─ h2o
│  ├─ obj.pkl
│  └─ object
│     ├─ book
│     ├─ cappuccino
│     ├─ chips
│     ├─ cocoa
│     ├─ espresso
│     ├─ lotion
│     ├─ milk
│     └─ spray
|
├─ grab
│  ├─ obj.pkl
│  ├─ processed_object_meshes
│  └─ contact_meshes
│     ├─ airplane.ply
│     ├─ alarmclock.ply
│     ├─ apple.ply
│     ├─ banana.ply
│     ...
|
├─ arctic
│  ├─ obj.pkl
│  └─ downloads/data/meta/object_vtemplates
│     ├─ box
│     ├─ capsulemachine
│     ├─ espressomachine
│     ├─ ketchup
│     ...
│
└─ mano
   └─ mano_v1_2
      └─ models
         ├─ info.txt
         ├─ LICENSE.txt
         ├─ MANO_LEFT.pkl
         └─ MANO_RIGHT.pkl

Preprocessing GRAB object

GRAB objects have so many vertiecs. So we reduce the number of vertices to 4,000.

python preprocessing_grab_object.py

Checkpoints (Coming soon)

Installation (Coming soon)

source scripts/install.sh

Demo (Coming soon)

source scripts/install.sh

For Future Work

Text description for Arctic

text2hoi's Issues

demo

Excellent work!! When will u show the visualization demo?

Code Release

Thank you for your intriguing work! Do you have an estimated timeline for when the code will be released