orientedformer

(TGRS 2024) OrientedFormer: An End-to-End Transformer-Based Oriented Object Detector in Remote Sensing Images

https://github.com/wokaikaixinxin/orientedformer

Science Score: 49.0%

This score indicates how likely this project is to be science-related based on various indicators:

○
CITATION.cff file
✓
codemeta.json file
Found codemeta.json file
✓
.zenodo.json file
Found .zenodo.json file
✓
DOI references
Found 2 DOI reference(s) in README
✓
Academic publication links
Links to: arxiv.org, ieee.org
○
Academic email domains
○
Institutional organization owner
○
JOSS paper metadata
○
Scientific vocabulary similarity
Low similarity (7.7%) to scientific vocabulary

Keywords

end-to-end object-detection oriented-object-detection remote-sensing transformer

Last synced: 6 months ago · JSON representation

Repository

(TGRS 2024) OrientedFormer: An End-to-End Transformer-Based Oriented Object Detector in Remote Sensing Images

Basic Info

Host: GitHub
Owner: wokaikaixinxin
License: apache-2.0
Language: Python
Default Branch: main
Homepage:
Size: 10.2 MB

Statistics

Stars: 35
Watchers: 2
Forks: 3
Open Issues: 3
Releases: 0

Topics

end-to-end object-detection oriented-object-detection remote-sensing transformer

Created over 1 year ago · Last pushed 8 months ago

Metadata Files

Readme Contributing License Citation

(TGRS 2024) OrientedFormer: An End-to-End Transformer-Based Oriented Object Detector in Remote Sensing Images

Please refer ai4rs ! Our ai4rs integrate remote sensing related work, including newest work on cvpr 2025.

The Chinese Version is below (中文版在下面).

Introduction

The paper is officially accepted by IEEE Transactions on Geoscience and Remote Sensing (TGRS 2024).

TGRS paper link https://ieeexplore.ieee.org/document/10669376

arxiv link https://arxiv.org/abs/2409.19648

NEW

✅ ICDAR2015 Dataset in MMRotate-1.x

✅ ICDAR2015 Metric in MMRotate-1.x

✅ ChannelMapperWithGN in MMRotate-1.x

✅ RBBoxL1Cost in MMRotate-1.x

✅ RotatedIoUCost in MMRotate-1.x

✅ TopkHungarianAssigner in MMRotate-1.x

If you like it, please click on star.

Installation

Please refer to Installation for more detailed instruction.

Note: Our codes base on the newest version mmrotate-1.x, not mmrotate-0.x.

Note: All of our codes can be found in path './projects/OrientedFormer/'.

Data Preparation

DOTA and DIOR-R : Please refer to Preparation for more detailed data preparation.

ICDAR2015 : (1) Download ICDAR2015 dataset from official link. (2) The data structure is as follows:

bash root ├── icdar2015 │ ├── ic15_textdet_train_img │ ├── ic15_textdet_train_gt │ ├── ic15_textdet_test_img │ ├── ic15_textdet_test_gt

Train

1). DIOR-R

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dior.py 2

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_swin-tiny_q300_layer2_head64_point32_1x_dior.py 2

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_lsk_t_q300_layer2_head64_point32_1x_dior.py 2

2). DOTA-v1.0

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dotav1.0.py 2

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dotav1.0-ms.py 2

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_r101_q300_layer2_head64_point32_1x_dotav1.0.py 2

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_swin-tiny_q300_layer2_head64_point32_1x_dotav1.0.py 2

3). DOTA-v1.5

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dotav1.5.py 2

4). DOTA-v2.0

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dotav2.0.py 2

5). ICDAR2015

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_r50_q300_layer2_head64_point32_2x_icdar2015.py 2

Test

1). DIOR-R

bash bash tools/dist_test.sh projects/OrientedFormer/configs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dior.py work_dirs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dior/epoch_12.pth 2

bash bash tools/dist_test.sh projects/OrientedFormer/configs/orientedformer_le90_swin-tiny_q300_layer2_head64_point32_1x_dior.py work_dirs/orientedformer_le90_swin-tiny_q300_layer2_head64_point32_1x_dior/epoch_12.pth 2

bash bash tools/dist_test.sh projects/OrientedFormer/configs/orientedformer_le90_lsk_t_q300_layer2_head64_point32_1x_dior.py work_dirs/orientedformer_le90_lsk_t_q300_layer2_head64_point32_1x_dior/epoch_12.pth 2

2). DOTA-v1.0

bash bash tools/dist_test.sh projects/OrientedFormer/configs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dotav1.0.py work_dirs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dotav1.0/epoch_12.pth 2

bash bash tools/dist_test.sh projects/OrientedFormer/configs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dotav1.0-ms.py work_dirs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dotav1.0-ms/epoch_12.pth 2

bash bash tools/dist_test.sh projects/OrientedFormer/configs/orientedformer_le90_r101_q300_layer2_head64_point32_1x_dotav1.0.py work_dirs/orientedformer_le90_r101_q300_layer2_head64_point32_1x_dotav1.0/epoch_12.pth 2

bash bash tools/dist_test.sh projects/OrientedFormer/configs/orientedformer_le90_swin-tiny_q300_layer2_head64_point32_1x_dotav1.0.py work_dirs/orientedformer_le90_swin-tiny_q300_layer2_head64_point32_1x_dotav1.0/epoch_12.pth 2

Upload results to DOTA official website.

3). DOTA-v1.5

bash bash tools/dist_test.sh projects/OrientedFormer/configs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dotav1.5.py work_dirs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dotav1.5/epoch_12.pth 2

Upload results to DOTA official website.

4). DOTA-v2.0

bash bash tools/dist_test.sh projects/OrientedFormer/configs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dotav2.0.py work_dirs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dotav2.0/epoch_12.pth 2

Upload results to DOTA official website.

5). ICDAR2015

Get result submit.zip

python tools/test.py projects/OrientedFormer/configs/orientedformer_le90_r50_q300_layer2_head64_point32_2x_icdar2015.py work_dirs/orientedformer_le90_r50_q300_layer2_head64_point32_2x_icdar2015/epoch_21.pth

Calculate precision, recall and F-measure. The script.py adapted from official website.

pip install Polygon3 python projects/icdar2015_evaluation/script.py –g=gt.zip –s=submit.zip

Main Result

1). DOTA-v1.0

| Dataset | Configs | Download | AP_50 | AP75 | mAP | Backbone | lr schd | bs | | --------- | -------------------- | ------------------------------------------------------------ | ------------ | --------- | ----------- | -------- | ------- | ---------------- | | DOTA-v1.0 | orientedformerle90r50q300layer2head64point321xdotav1.0.py | Hugging Face | 75.3729 | 46.390216 | 45.0071 | R50 | 12epoch | 2img2 rtx2080ti | | DOTA-v1.0 | orientedformerle90r50q300layer2head64point321xdotav1.0-ms.py | Hugging Face | 79.064371 | 57.463 | 51.891899 | R50 | 12epoch | 2img2 rtx2080ti | | DOTA-v1.0 | orientedformerle90r101q300layer2head64point321xdotav1.0.py | Hugging Face | 75.915958978 | 49.76108 | 47.11829758 | R101 | 12epoch | 2img2 rtx2080ti | | DOTA-v1.0 | orientedformerle90swin-tinyq300layer2head64point321xdotav1.0.py | Hugging Face | 75.8819 | 48.965 | 45.8218 | Swin-T | 12epoch | 2img2 rtx2080ti |

2). DOTA-v1.5

| Dataset | Configs | Download | AP_50 | Backbone | lr schd | bs | | --------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ----- | -------- | ------- | ---------------- | | DOTA-v1.5 | orientedformerle90r50q300layer2head64point321xdotav1.5.py | Hugging Face | 67.06 | R50 | 12epoch | 2img*2 rtx2080ti |

Due to the limitation of the length of the paper, all categories of AP for DOTA-1.5 are not available in the paper. Here is a list:

| PL | BD | BR | GTF | SV | LV | SH | TC | BC | ST | SBF | RA | HA | SP | HC | CC | AP50 | AP75 | mAP | | ------- | ------- | ------- | ------- | ------- | ------- | ------- | -------- | -------- | --------- | --------- | ------ | --------- | --------- | ------- | --------- | -------- | ------- | -------- | | 72.0444 | 77.4554 | 51.2471 | 64.9538 | 64.0453 | 77.0387 | 85.3310 | 90.83699 | 77.31017 | 78.106886 | 56.103059 | 68.776 | 68.140988 | 72.081567 | 58.6135 | 10.855397 | 67.05879 | 39.2845 | 38.78675 |

3). DOTA-v2.0

| Dataset | Configs | Download | AP_50 | Backbone | lr schd | bs | | --------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ----- | -------- | ------- | ---------------- | | DOTA-v2.0 | orientedformerle90r50q300layer2head64point321xdotav2.0.py | Hugging Face | 54.27 | R50 | 12epoch | 2img*2 rtx2080ti |

Due to the limitation of the length of the paper, all categories of AP for DOTA-2.0 are not available in the paper. Here is a list:

| PL | BD | BR | GTF | SV | LV | SH | TC | BC | ST | SBF | RA | HA | SP | HC | CC | airport | helipad | AP50 | AP75 | mAP | | ------- | -------- | ---------- | --------- | --------- | -------- | --------- | ------- | ------- | ------- | ------- | --------- | ------- | -------- | -------- | ------ | -------- | ------- | --------- | ---------- | ---------- | | 76.7619 | 51.55655 | 42.3872759 | 60.464159 | 56.482355 | 55.43076 | 66.681058 | 78.6341 | 60.0626 | 69.6894 | 35.0316 | 56.015956 | 51.9962 | 56.20235 | 54.95597 | 24.335 | 67.31572 | 12.9641 | 54.266644 | 28.8561385 | 30.0281367 |

4). DIOR-R

| Dataset | Configs | Download | AP_50 | Backbone | lr schd | bs | | ------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ----- | -------- | ------- | ---------------- | | DIOR-R | orientedformerle90r50q300layer2head64point321xdior.py | Hugging Face | 67.28 | R50 | 12epoch | 2img2 rtx2080ti | | DIOR-R | orientedformerle90swin-tinyq300layer2head64point321xdior.py | Hugging Face | 68.84 | Swin-T | 12epoch | 2img2 rtx2080ti | | DIOR-R | orientedformerle90lsktq300layer2head64point321x_dior.py | Hugging Face | 65.07 | LSK-Net | 12epoch | 2img*2 rtx2080ti |

5). ICDAR-2015

| Dataset | Configs | Download | P | R | F-measure | Backbone | lr schd | bs | | --------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ---- | ---- | --------- | -------- | ------- | ---------------- | | ICDAR2015 | orientedformerle90r50q300layer2head64point322xicdar2015.py | Hugging Face | 85.3 | 74.2 | 79.4 | R50 | 24epoch | 2img*2 rtx2080ti |

Cite OrientedFormer

@ARTICLE{10669376, author={Zhao, Jiaqi and Ding, Zeyu and Zhou, Yong and Zhu, Hancheng and Du, Wen-Liang and Yao, Rui and El Saddik, Abdulmotaleb}, journal={IEEE Transactions on Geoscience and Remote Sensing}, title={OrientedFormer: An End-to-End Transformer-Based Oriented Object Detector in Remote Sensing Images}, year={2024}, volume={62}, number={}, pages={1-16}, keywords={Encoding;Object detection;Proposals;Detectors;Remote sensing;Current transformers;Position measurement;End-to-end detectors;oriented object detection;positional encoding (PE);remote sensing;transformer}, doi={10.1109/TGRS.2024.3456240}}

(TGRS 2024) OrientedFormer: An End-to-End Transformer-Based Oriented Object Detector in Remote Sensing Images

请参考 ai4rs ! 我们的ai4rs整合了最新遥感相关工作，包括最新的CVPR2025的工作。

简介

论文被IEEE Transactions on Geoscience and Remote Sensing (TGRS 2024) 接受。

TGRS官方论文链接 https://ieeexplore.ieee.org/document/10669376

arxiv link https://arxiv.org/abs/2409.19648

新特性：

✅ 数据集工具： ICDAR2015 Dataset in MMRotate-1.x

✅ 数据集工具： ICDAR2015 Metric in MMRotate-1.x

✅ 不用FPN，用channelmapper： ChannelMapperWithGN in MMRotate-1.x

✅ 端到端分配的L1代价矩阵： RBBoxL1Cost in MMRotate-1.x

✅ 端到端分配的IoU代价矩阵： RotatedIoUCost in MMRotate-1.x

✅ Topk匈牙利匹配： TopkHungarianAssigner in MMRotate-1.x

如果喜欢，请点一点小星星收藏。

安装

参考mmrotate-1.x的官方安装教程获取更多安装细节。

注意：代码是基于最新版本的mmrotate-1.x，而不是旧版的mmrotate-0.x。

注意：orientedformer的代码全部在路径'./projects/OrientedFormer/'

数据准备

DOTA and DIOR-R : 请参考mmrotate-1.x官方数据处理对DOTA和DIOR-R的准备方法。

ICDAR2015 : (1) 从官方网站下载 ICDAR2015 数据集。(2) 数据路径结构如下所示：

bash root ├── icdar2015 │ ├── ic15_textdet_train_img │ ├── ic15_textdet_train_gt │ ├── ic15_textdet_test_img │ ├── ic15_textdet_test_gt

训练

1). DIOR-R

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dior.py 2

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_swin-tiny_q300_layer2_head64_point32_1x_dior.py 2

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_lsk_t_q300_layer2_head64_point32_1x_dior.py 2

2). DOTA-v1.0

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dotav1.0.py 2

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dotav1.0-ms.py 2

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_r101_q300_layer2_head64_point32_1x_dotav1.0.py 2

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_swin-tiny_q300_layer2_head64_point32_1x_dotav1.0.py 2

3). DOTA-v1.5

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dotav1.5.py 2

4). DOTA-v2.0

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_r50_q300_layer2_head64_point32_1x_dotav2.0.py 2

5). ICDAR2015

bash bash tools/dist_train.sh projects/OrientedFormer/configs/orientedformer_le90_r50_q300_layer2_head64_point32_2x_icdar2015.py 2

测试

1). DIOR-R

2). DOTA-v1.0

将结果上传 DOTA官方网站。

3). DOTA-v1.5

将结果上传 DOTA官方网站。

4). DOTA-v2.0

将结果上传 DOTA官方网站。

5). ICDAR2015

得到结果submit.zip

计算precision, recall 和 F-measure

pip install Polygon3 python projects/icdar2015_evaluation/script.py –g=gt.zip –s=submit.zip

主要结果

1). DOTA-v1.0

| Dataset | Configs | Download | AP_50 | AP75 | mAP | Backbone | lr schd | bs | | --------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------ | --------- | ----------- | -------- | ------- | ---------------- | | DOTA-v1.0 | orientedformerle90r50q300layer2head64point321xdotav1.0.py | Hugging Face | 75.3729 | 46.390216 | 45.0071 | R50 | 12epoch | 2img2 rtx2080ti | | DOTA-v1.0 | orientedformerle90r50q300layer2head64point321xdotav1.0-ms.py | Hugging Face | 79.064371 | 57.463 | 51.891899 | R50 | 12epoch | 2img2 rtx2080ti | | DOTA-v1.0 | orientedformerle90r101q300layer2head64point321xdotav1.0.py | Hugging Face | 75.915958978 | 49.76108 | 47.11829758 | R101 | 12epoch | 2img2 rtx2080ti | | DOTA-v1.0 | orientedformerle90swin-tinyq300layer2head64point321xdotav1.0.py | Hugging Face | 75.8819 | 48.965 | 45.8218 | Swin-T | 12epoch | 2img2 rtx2080ti |

2). DOTA-v1.5

由于论文长度的限制，论文中没有DOTA-1.5的所有类别的AP，这里列出：

3). DOTA-v2.0

| Dataset | Configs | Download | AP_50 | Backbone | lr schd | bs | | --------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ----- | -------- | ------- | ---------------- | | DOTA-v2.0 | orientedformerle90r50q300layer2head64point321xdotav2.0.py | Hugging Face | 54.27 | R50 | 12epoch | 2img*2 rtx2080ti |

由于论文长度的限制，论文中没有DOTA-2.0的所有类别的AP，这里列出：

4). DIOR-R

5). ICDAR-2015

| Dataset | Configs | Download | P | R | F-measure | Backbone | lr schd | bs | | --------- | --------------------------------|-----------|----------------- | ------------------------------------------------------------ | ---- | -------- | ------- | ---------------- | | ICDAR2015 | orientedformerle90r50q300layer2head64point322xicdar2015.py | Hugging Face | 85.3 | 74.2 | 79.4 | R50 |24epoch|2img*2 rtx2080ti|

引用 OrientedFormer

Owner

Login: wokaikaixinxin
Kind: user

Repositories: 1
Profile: https://github.com/wokaikaixinxin

GitHub Events

Total

Issues event: 4
Watch event: 24
Issue comment event: 14
Push event: 8
Fork event: 2

Last Year

Issues event: 4
Watch event: 24
Issue comment event: 14
Push event: 8
Fork event: 2

Dependencies

.github/workflows/lint.yml actions

actions/checkout v2 composite
actions/setup-python v2 composite

.github/workflows/merge_stage_test.yml actions

actions/checkout v2 composite
actions/setup-python v2 composite
codecov/codecov-action v1.0.14 composite

.github/workflows/pr_stage_test.yml actions

actions/checkout v2 composite
actions/setup-python v2 composite

.github/workflows/publish-to-pypi.yml actions

actions/checkout v2 composite
actions/setup-python v1 composite

.github/workflows/test_mim.yml actions

actions/checkout v2 composite
actions/setup-python v2 composite

.circleci/docker/Dockerfile docker

pytorch/pytorch ${PYTORCH}-cuda${CUDA}-cudnn${CUDNN}-devel build

docker/Dockerfile docker

pytorch/pytorch ${PYTORCH}-cuda${CUDA}-cudnn${CUDNN}-devel build

docker/serve/Dockerfile docker

pytorch/pytorch ${PYTORCH}-cuda${CUDA}-cudnn${CUDNN}-devel build

requirements/build.txt pypi

cython *
numpy *

requirements/docs.txt pypi

docutils ==0.16.0
myst-parser *
sphinx ==4.0.2
sphinx-copybutton *
sphinx_markdown_tables *
sphinx_rtd_theme ==0.5.2

requirements/mminstall.txt pypi

mmcv >=2.0.0rc2,<2.1.0
mmdet >=3.0.0rc2,<3.1.0
mmengine >=0.1.0

requirements/optional.txt pypi

imagecorruptions *
scikit-learn *
scipy *

requirements/readthedocs.txt pypi

e2cnn *
mmcv >=2.0.0rc2
mmdet >=3.0.0rc2
mmengine >=0.1.0
torch *
torchvision *

requirements/runtime.txt pypi

matplotlib *
numpy *
pycocotools *
six *
terminaltables *
torch *

requirements/tests.txt pypi

asynctest * test
codecov * test
coverage * test
cython * test
flake8 * test
interrogate * test
isort ==4.3.21 test
kwarray * test
matplotlib * test
parameterized * test
pytest * test
scikit-learn * test
ubelt * test
wheel * test
xdoctest >=0.10.0 test
yapf * test

requirements.txt pypi

setup.py pypi

ecosyste.ms

Data

Tools

Indexes

Applications

Experiments

Open Source Science

orientedformer

Science Score: 49.0%

Keywords

Repository

Basic Info

Statistics

Topics

Metadata Files

README.md

(TGRS 2024) OrientedFormer: An End-to-End Transformer-Based Oriented Object Detector in Remote Sensing Images

Introduction

Installation

Data Preparation

Train

Test

Main Result

Cite OrientedFormer

(TGRS 2024) OrientedFormer: An End-to-End Transformer-Based Oriented Object Detector in Remote Sensing Images

简介

安装

数据准备

训练

测试

主要结果

引用 OrientedFormer

Owner

GitHub Events

Total

Last Year

Dependencies