MGA-Net
We provide PyTorch implementation for "Modular Graph Attention Network for Complex Visual Relational Reasoning". (ACCV2020)
Paper
- Modular Graph Attention Network for Complex Visual Relational Reasoning
- Yihan Zheng *, Zhiquan Wen *, Mingkui Tan, Runhao Zeng, Qi Chen, Yaowei Wang * and Qi Wu
- Asian Conference on Computer Vision (ACCV), 2020
Dependencies
- Python 3.6
- requirements.txt
Getting Started
Installation
-
Clone this repository:
git clone https://github.com/Zhiquan-Wen/MGA-Net.git cd MGA-Net export PYTHONPATH='./' -
Install PyTorch and other dependencies:
pip install -r requirements.txt
Training
-
Download the features to the folder "data" from the releases.
wget https://github.com/Zhiquan-Wen/MGA-Net/releases/download/feature_v1/clevr_ref+_1.0.zip unzip -d ./data clevr_ref+_1.0.zip -
Train the MGA-Net
python train.py
Pre-trained Model
| Model | GGNN | Link | Acc (%) |
|---|---|---|---|
| MGA-Net | 0 | https://github.com/Zhiquan-Wen/MGA-Net/releases/download/Pretrained_model/GGNN_0.pth | 76.51 |
| MGA-Net | 3 | https://github.com/Zhiquan-Wen/MGA-Net/releases/download/Pretrained_model/GGNN_3.pth | 80.87 |
Citation
If this work is useful for your research, please cite our paper:
@inproceedings{MGA-Net,
author = {Yihan Zheng and
Zhiquan Wen and
Mingkui Tan and
Runhao Zeng and
Qi Chen and
Yaowei Wang and
Qi Wu},
title = {Modular Graph Attention Network for Complex Visual Relational Reasoning},
booktitle = {ACCV},
pages = {137--153},
year = {2020},
}