https://github.com/njvisionpower/mtcnn-train-gluon
Implement MTCNN train pipeline with MXNet gluon
https://github.com/njvisionpower/mtcnn-train-gluon
face-detection gluon mxnet train-mtcnn
Last synced: 5 months ago
JSON representation
Implement MTCNN train pipeline with MXNet gluon
- Host: GitHub
- URL: https://github.com/njvisionpower/mtcnn-train-gluon
- Owner: njvisionpower
- Created: 2019-11-09T13:20:02.000Z (about 6 years ago)
- Default Branch: master
- Last Pushed: 2020-10-02T07:20:32.000Z (about 5 years ago)
- Last Synced: 2025-06-25T13:02:22.325Z (5 months ago)
- Topics: face-detection, gluon, mxnet, train-mtcnn
- Language: Python
- Size: 13.3 MB
- Stars: 4
- Watchers: 1
- Forks: 3
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# mtcnn-train-gluon
Implement MTCNN train pipeline with MXNet gluon
## Introduction
This repo mainly implement MTCNN([Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Networks](https://arxiv.org/ftp/arxiv/papers/1604/1604.02878.pdf)) with mxnet gluon framework. I train a faster network instead of original version(only replace pooling with conv + stride), just design your own network and you may get faster and more accurate face detection model.
## How to run
```
python MTCNN.py
```
## How to train
The WIDER-FACE dataset is needed, download and unzip to datset directory such as:
```
---dataset
---WIDER_FACE
---wider_face_split
---WIDER_train
---WIDER_val
```
First train PNet, then RNet and finally ONet. The dataset process part reference at [repo](https://github.com/beichen2012/mtcnn-pytorch). Steps list:
**1. Generate PNet data**
```
python generate_data\Generate_PNet_data.py
```
**2. Train PNet**
```
python train_pnet.py
```
**3. Generate RNet data with trained PNet**
```
python generate_data\Generate_RNet_data.py
```
**4. Train RNet**
```
python train_rnet.py
```
**5. Generate ONet data with trained PNet and RNet**
```
python generate_data\Generate_ONet_data.py
```
**6. Train ONet**
```
python train_onet.py
```
## Demo
## Reference
[Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Networks](https://arxiv.org/ftp/arxiv/papers/1604/1604.02878.pdf)
[mtcnn-pytorch](https://github.com/beichen2012/mtcnn-pytorch)