Skip to content

A Simple Version of GraphFPN: Retinanet using Graph feature pyramid networks

License

Notifications You must be signed in to change notification settings

GangmingZhao/Graph-FPN-Retina

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

89 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

GraphFPN: Graph Feature Pyramid Network for Object Detection with Retinanet

Inspired by Graph_FPN and RetinaNet, we have used the Graph_FPN structure as a backbone to train on Retinanet and the work is not yet complete.

Currently working with the author Gangming Zhao, we will improve our code on mmdetection in the future.

For demonstraction with Graph_fpn, run:

python demo.py

For demonstraction with fpn, run:

python demo.py --no_graph

For training , run:

python train.py

For test with Graph_fpn, run

python test.py

For test with fpn, run

python test.py --no_graph

If You need COCO API for test, you can download from here. You need to set the backend of DGL to tensorflow, here is tutorial link

Folder structure

${ROOT}
└── checkpoint/
└── COCO/    
│   └── coco/
│   │    ├── .config 
│   │    ├── 2017/
│   │
│   ├── downloads/
│
│
└── data_demo/
|   ├── data/
|   |    ├── coco
|   |    ├── checkpoint
|   ├── data.zip
|
├── results/
├── src/     
|   ├── configs/
|   |    ├── configs.py
|   |
|   ├── detection/
|   |    ├── datasets/
|   |    |      ├── coco.py
|   |    ├── utils/
|   |
|   ├── model/
|   ├── init_path.py
|   ├── demo.py
|   ├── train.py
|   ├── test.py
├── README.md 
└── requirements.txt

References

[1] Retinanet: Focal Loss for Dense Object Detection
[2] Graph-FPN: GraphFPN: Graph Feature Pyramid Network for Object Detection
[3] Object Detection with RetinaNet: Keras Implementation

About

A Simple Version of GraphFPN: Retinanet using Graph feature pyramid networks

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%