Towards Fine-Grained HBOE with Rendered Orientation Set and Laplace Smoothing
Ruisi Zhao, Mingming Li, Zheng Yang, Binbin Lin, Xiaohui Zhong, Xiaobo Ren, Deng Cai, Boxi Wu
Official codes for Towards Fine-Grained HBOE. This repository contains two version of implementation.
- Original version (Paper results, evaluation)
- Optimized version (Less memory usage, simpler model. Please see the code for details, specifically in the model section, we design a new head with local-window attention. Compared to OEFormer, this version model significantly reduces training costs. We show the results in the following image.)
- [√] Code Released.
- [√] Checkpoint Released.
- [√] Data Released.
-
Prepare environment
conda create -n hboe python=3.9 -y conda activate hboe pip install -r requirements.txt
-
Clone repo & install
git clone https://github.com/Whalesong-zrs/Towards-Fine-grained-HBOE.git cd towards-fine-grained-hboe
Download the checkpoints and put them in correct paths.
pretrained ckpt:
pretrained_models/pretrained_hrnet.pth
pretrained_models/pretrained_oeformer.pth
trained ckpt:
checkpoints/hrnet_head.pth
checkpoints/oeformer.pth
If you want to quickly reimplement our methods, we provide the following resources used in the paper. For MEBOW dataset, please go to MEBOW for more details.
Paper Resources | Rendered Datasets | Checkpoints |
---|---|---|
Download Link | Google Drive | Google Drive |
After downloading, the path should be arranged as follows:
towards-fine-grained-hboe
├── checkpoints
| ├── hrnet_head.pth # Trained checkpoints
│ ├── oeformer.pth
├── experiments
| ├── hrnet_head.yaml # Training / Inference setting
| ├── oeformer.yaml
├── imgs
├── lib
│ ├── config # Some scripts about config
│ ├── core # Train and enaluate
│ ├── dataset
│ ├── utils
├── logs
├── model_arch
├── output
├── pretrained_models
├── tools
...
# check your config
bash train.sh
# check your config
bash test.sh
This codebase builds on MEBOW. Thanks for open-sourcing! Besides, we acknowledge following great open-sourcing projects:
@inproceedings{zhao2024towards,
title={Towards Fine-Grained HBOE with Rendered Orientation Set and Laplace Smoothing},
author={Zhao, Ruisi and Li, Mingming and Yang, Zheng and Lin, Binbin and Zhong, Xiaohui and Ren, Xiaobo and Cai, Deng and Wu, Boxi},
booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
volume={38},
number={7},
pages={7505--7513},
year={2024}
}
If you have any questions and improvement suggestions, please email Ruisi Zhao (zhaors00@zju.edu.cn), or open an issue.