JDet

Introduction

JDet is an object detection benchmark based on Jittor, and mainly focus on aerial image object detection (oriented object detection).

Install

JDet environment requirements:

• System: Linux(e.g. Ubuntu/CentOS/Arch), macOS, or Windows Subsystem of Linux (WSL)
• Python version >= 3.7
• CPU compiler (require at least one of the following)
  • g++ (>=5.4.0)
  • clang (>=8.0)
• GPU compiler (optional)
  • nvcc (>=10.0 for g++ or >=10.2 for clang)
• GPU library: cudnn-dev (recommend tar file installation, reference link)

Step 1: Install the requirements

git clone https://github.com/Jittor/JDet
cd JDet
python -m pip install -r requirements.txt

If you have any installation problems for Jittor, please refer to Jittor

Step 2: Install JDet

cd JDet
# suggest this 
python setup.py develop
# or
python setup.py install

If you don't have permission for install,please add --user.

Or use PYTHONPATH: You can add export PYTHONPATH=$PYTHONPATH:{you_own_path}/JDet/python into .bashrc, and run

source .bashrc

Getting Started

Datasets

The following datasets are supported in JDet, please check the corresponding document before use.

DOTA1.0/DOTA1.5/DOTA2.0 Dataset: dota.md.

FAIR Dataset: fair.md

SSDD/SSDD+: ssdd.md

You can also build your own dataset by convert your datas to DOTA format.

Config

JDet defines the used model, dataset and training/testing method by config-file, please check the config.md to learn how it works.

Train

python tools/run_net.py --config-file=configs/s2anet_r50_fpn_1x_dota.py --task=train

Test

If you want to test the downloaded trained models, please set resume_path={you_checkpointspath} in the last line of the config file.

python tools/run_net.py --config-file=configs/s2anet_r50_fpn_1x_dota.py --task=test

Test on images / Visualization

You can test and visualize results on your own image sets by:

python tools/run_net.py --config-file=configs/s2anet_r50_fpn_1x_dota.py --task=vis_test

You can choose the visualization style you prefer, for more details about visualization, please refer to visualization.md.

Build a New Project

In this section, we will introduce how to build a new project(model) with JDet. We need to install JDet first, and build a new project by:

mkdir $PROJECT_PATH$
cd $PROJECT_PATH$
cp $JDet_PATH$/tools/run_net.py ./
mkdir configs

Then we can build and edit configs/base.py like $JDet_PATH$/configs/retinanet.py. If we need to use a new layer, we can define this layer at $PROJECT_PATH$/layers.py and import layers.py in $PROJECT_PATH$/run_net.py, then we can use this layer in config files. Then we can train/test this model by:

python run_net.py --config-file=configs/base.py --task=train
python run_net.py --config-file=configs/base.py --task=test

Models

Models	Dataset	Sub_Image_Size/Overlap	Train Aug	Test Aug	Optim	Lr schd	mAP	Paper	Config	Download
S2ANet-R50-FPN	DOTA1.0	1024/200	flip	-	SGD	1x	74.11	arxiv	config	model
S2ANet-R50-FPN	DOTA1.0	1024/200	flip+ra90+bc	-	SGD	1x	76.40	arxiv	config	model
S2ANet-R50-FPN	DOTA1.0	1024/200	flip+ra90+bc+ms	ms	SGD	1x	79.72	arxiv	config	model
S2ANet-R101-FPN	DOTA1.0	1024/200	Flip	-	SGD	1x	74.28	arxiv	config	model
Gliding-R50-FPN	DOTA1.0	1024/200	Flip	-	SGD	1x	72.93	arxiv	config	model
Gliding-R50-FPN	DOTA1.0	1024/200	Flip+ra90+bc	-	SGD	1x	74.93	arxiv	config	model
H2RBox-R50-FPN	DOTA1.0	1024/200	flip	-	AdamW	1x	67.62	arxiv	config	model
RetinaNet-hbb-R50-FPN	DOTA1.0	1024/200	flip	-	SGD	1x	68.02	arxiv	config	model
RetinaNet-obb-R50-FPN	DOTA1.0	1024/200	flip	-	SGD	1x	68.07	arxiv	config	model
GWD-R50-FPN	DOTA1.0	1024/200	flip	-	SGD	1x	68.88	arxiv	config	model
KLD-R50-FPN	DOTA1.0	1024/200	flip	-	SGD	1x	69.10	arxiv	config	model
KFIoU-R50-FPN	DOTA1.0	1024/200	flip	-	SGD	1x	69.36	arxiv	config	model
FasterRCNN-R50-FPN	DOTA1.0	1024/200	Flip	-	SGD	1x	69.631	arxiv	config	model
RoITransformer-R50-FPN	DOTA1.0	1024/200	Flip	-	SGD	1x	73.842	arxiv	config	model
FCOS-R50-FPN	DOTA1.0	1024/200	flip	-	SGD	1x	70.40	ICCV19	config	model
OrientedRCNN-R50-FPN	DOTA1.0	1024/200	Flip	-	SGD	1x	75.62	ICCV21	config	model
ReDet-R50-FPN	DOTA1.0	1024/200	Flip	-	SGD	1x	76.23	arxiv	config	model pretrained
CSL-R50-FPN	DOTA1.0	1024/200	flip	-	SGD	1x	67.99	arxiv	config	model
RSDet-R50-FPN	DOTA1.0	1024/200	Flip	-	SGD	1x	68.41	arxiv	config	model
ATSS-R50-FPN	DOTA1.0	1024/200	flip	-	SGD	1x	72.44	arxiv	config	model
Reppoints-R50-FPN	DOTA1.0	1024/200	flip	-	SGD	1x	56.34	arxiv	config	model

Notice:

1. ms: multiscale
2. flip: random flip
3. ra: rotate aug
4. ra90: rotate aug with angle 90,180,270
5. 1x : 12 epochs
6. bc: balance category
7. mAP: mean Average Precision on DOTA1.0 test set

Plan of Models

✔️Supported 🕒Doing ➕TODO

• ✔️ S2ANet
• ✔️ Gliding
• ✔️ RetinaNet
• ✔️ Rotated RetinaNet
• ✔️ Faster R-CNN
• ✔️ SSD
• ✔️ ROI Transformer
• ✔️ FCOS
• ✔️ Oriented R-CNN
• ✔️ YOLOv5
• ✔️ GWD
• ✔️ KLD
• ✔️ H2RBox
• ✔️ KFIoU
• ✔️ Localization Distillation
• ✔️ ReDet
• ✔️ CSL
• ✔️ Reppoints
• ✔️ RSDet
• ✔️ ATSS
• 🕒 R3Det
• 🕒 Cascade R-CNN
• 🕒 Oriented Reppoints
• ➕ DCL
• ➕ Double Head OBB
• ➕ Guided Anchoring
• ➕ Sampling Equivariant Self-attention Networks
• ➕ ...

Plan of Datasets

✔️Supported 🕒Doing ➕TODO

• ✔️ DOTA1.0
• ✔️ DOTA1.5
• ✔️ DOTA2.0
• ✔️ SSDD
• ✔️ SSDD+
• ✔️ FAIR
• ✔️ COCO
• ➕ LS-SSDD
• ➕ DIOR-R
• ➕ HRSC2016
• ➕ ICDAR2015
• ➕ ICDAR2017 MLT
• ➕ UCAS-AOD
• ➕ FDDB
• ➕ OHD-SJTU
• ➕ MSRA-TD500
• ➕ Total-Text
• ➕ ...

Contact Us

Website: http://cg.cs.tsinghua.edu.cn/jittor/

Email: [email protected]

File an issue: https://github.com/Jittor/jittor/issues

QQ Group: 761222083

The Team

JDet is currently maintained by the Tsinghua CSCG Group. If you are also interested in JDet and want to improve it, Please join us!

Citation

@article{hu2020jittor,
  title={Jittor: a novel deep learning framework with meta-operators and unified graph execution},
  author={Hu, Shi-Min and Liang, Dun and Yang, Guo-Ye and Yang, Guo-Wei and Zhou, Wen-Yang},
  journal={Science China Information Sciences},
  volume={63},
  number={222103},
  pages={1--21},
  year={2020}
}

Reference

1. Jittor
2. mmrotate
3. Detectron2
4. mmdetection
5. maskrcnn_benchmark
6. RotationDetection
7. s2anet
8. gliding_vertex
9. oriented_rcnn
10. r3det
11. AerialDetection
12. DOTA_devkit
13. OBBDetection