https://github.com/JJXiangJiaoJun/gluon_PyramidBox
PyramidBox implemented with mxnet
https://github.com/JJXiangJiaoJun/gluon_PyramidBox
gluon mxnet paper pyramidbox
Last synced: 22 days ago
JSON representation
PyramidBox implemented with mxnet
- Host: GitHub
- URL: https://github.com/JJXiangJiaoJun/gluon_PyramidBox
- Owner: JJXiangJiaoJun
- Created: 2019-01-26T16:05:02.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2019-07-21T02:16:58.000Z (almost 6 years ago)
- Last Synced: 2024-11-08T18:45:09.393Z (6 months ago)
- Topics: gluon, mxnet, paper, pyramidbox
- Language: Python
- Homepage:
- Size: 915 KB
- Stars: 8
- Watchers: 1
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
- Awesome-MXNet - PyramidBox
README
# PyramidBox: A Context-assisted Single Shot Face Detector
## Introduction
A MXNet implementation of PyramiBbox:A Context-assisted Single Shot Face Detector.
If you want to learn more details,please refer to the original paper.
```
@inproceedings{Tang2018PyramidBoxAC,
title={PyramidBox: A Context-Assisted Single Shot Face Detector},
author={Xu Tang and Daniel K. Du and Zeqiang He and Jingtuo Liu},
booktitle={ECCV},
year={2018}
}
```
I train PyramidBox with WIDER FACE dataset,results are as follows:
| |Easy mAP|Medium mAP|Hard mAP|
|---|---|---|---|
|paper|**96.1**|**95.0**|**88.9**|
|this repo|**92.5**|**90.8**|**83.3**|
I think mainly reasons that this repo can not get the same precision as paper as follows:
* I use batch size 4 because of memory limitations,which is 16 in the paper
* some parameters are not metioned in the paper
Here are several examples of succesful detection outputs:
## Details
I implement following structures metioned in the paper:
- [x] Low-Level FPN
- [x] max-in-out layer
- [x] PyramidAnchors
- [x] Context-sensitive Prediction Module
- [x] Data-anchor sampling
- [x] Learning rate warmup and cosine decay
## Dependencies
* Python 3.6.4
* [MXNet](https://github.com/apache/incubator-mxnet) 1.3.1
* [gluon-cv](https://github.com/dmlc/gluon-cv) 0.4.1
## Implement Details
* Ubuntu 16.04 LTS
* CUDA 9.0
* cuDNN 7.0.5
## Preparation
```bash
git clone [email protected]:JJXiangJiaoJun/gluon_PyramidBox.git
cd gluon_PyramidBox
```
## Download and prepare data
1. download [WIDER FACE](http://mmlab.ie.cuhk.edu.hk/projects/WIDERFace/) dataset into `widerface/downloads`
```bash
$$ tree widerface/downloads
widerface/downloads
├── eval_tools.zip
├── Submission_example.zip
├── wider_face_split.zip
├── WIDER_test.zip
├── WIDER_train.zip
└── WIDER_val.zip
```
2. Prepare data: unzip data, annotations and eval_tools to `./widerface`
```bash
python tool/prepare.py
$$ tree widerface -L 1
widerface
├── downloads
├── eval_tools
├── wider_face_split
├── WIDER_train
└── WIDER_val
```
3. Prepare custom val dataset for quick validation (crop and resize to 640)
```bash
$$ python tool/build_custom_val.py
$$ tree widerface -L 1
widerface
├── downloads
├── eval_tools
├── WIDER_custom
├── wider_face_split
├── WIDER_train
└── WIDER_val
```
## Train on WIDER FACE Datasets
train vgg16 based pyramidbox with 1 gpus.I only implement VGG16 as backbone currently:
```bash
python train_end2end.py --use-bn
```
or you can see more details:
```bash
python train_end2end.py --help
```
## Evalution
eval your own model on WIDER FACE Datasets:
```bash
python eval.py --use-bn --model models/pyramidbox/pyramidbox_best.params
```
## Reference
* [PyramidBox with pytorch](https://github.com/Goingqs/PyramidBox)
* [sfd.gluoncv](https://github.com/yangfly/sfd.gluoncv)