GREAT: Geometry-Intention Collaborative Inference for Open-Vocabulary 3D Object Affordance Grounding（CVPR 2025）

PyTorch implementation of GREAT: Geometry-Intention Collaborative Inference for Open-Vocabulary 3D Object Affordance Grounding. This repository contains PyTorch training, evaluation, inference code, pretrained models and PIADv2 dataset.

✅ TODO List

• Release paper on arxiv.
• Release the training, evaluation, and inference code.
• Release the pretrained checkpoint.
• Release PIADv2 dataset.

📋 Table of content

1. 💡 Overview
2. 📖 Dataset
3. ✏️ Usage
   1. Environment
   2. Training
   3. Evaluation
   4. Render
   5. Reasoning
4. 🔍 Citation

💡Overview

📖 Dataset

Properties of the PIADv2 dataset. (a) Extensive data examples from PIADv2, the red region in point clouds is the affordance annotation. (b) Category distribution in PIADv2. (c) Confusion matrix between affordance and object categories, where the horizontal axis represents object category and the vertical axis represents affordance category.(d) Ratio of images and point clouds in each affordance category.

Download the PIADv2 dataset from Google Drive or Baidu Pan (key: PIAD).

✏️ Usage

Environment

First clone this respository and create a conda environment, as follows:

git clone https://github.com/yawen-shao/GREAT_code.git
cd GREAT_code
conda create -n great python=3.9 -y

Then, install the other dependancies:

pip install -r requirements.txt

Training

To train the GREAT model, you can modify the training parameter in config/config_seen_GREAT.yaml and then run the following command:

bash train.sh

Evaluation

To evaluate the trained GREAT model, run the following command:

bash evalization.sh

Render

Refer to IAG-Net for the object visualization.

Reasoning

If you want to define your own prompts to reason about object/affordance knowledge, you can refer to the model/MHACoT.py.
For specific reasoning implementations and how to finetune your own affordance model, refer to InternVL.

🔍 Citation

@article{GREAT_Shao,
  title={GREAT: Geometry-Intention Collaborative Inference for Open-Vocabulary 3D Object Affordance Grounding},
  author={Shao, Yawen and Zhai, Wei and Yang, Yuhang and Luo, Hongchen and Cao, Yang and Zha, Zheng-Jun},
  journal={arXiv preprint arXiv:2411.19626},
  year={2024}
}

Acknowledgement

The code is built based on IAGNet. Thanks for the excellent open-source code!!