https://github.com/saicoco/Gluon-PSENet
mxnet-Gluon implementation of PSENet text detector (Shape Robust Text Detection with Progressive Scale Expansion Network)
https://github.com/saicoco/Gluon-PSENet
mxnet-gluon psenet text-detection
Last synced: 3 days ago
JSON representation
mxnet-Gluon implementation of PSENet text detector (Shape Robust Text Detection with Progressive Scale Expansion Network)
- Host: GitHub
- URL: https://github.com/saicoco/Gluon-PSENet
- Owner: saicoco
- License: gpl-3.0
- Created: 2019-02-28T08:25:36.000Z (about 6 years ago)
- Default Branch: master
- Last Pushed: 2019-06-17T09:00:24.000Z (almost 6 years ago)
- Last Synced: 2024-08-01T22:41:31.092Z (9 months ago)
- Topics: mxnet-gluon, psenet, text-detection
- Language: C++
- Homepage:
- Size: 24.3 MB
- Stars: 18
- Watchers: 2
- Forks: 6
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- Awesome-MXNet - PSENet
README
# Shape Robust Text Detection with Progressive Scale Expansion Network
A reimplement of PSENet with mxnet-gluon. Just train on ICPR.
- *Support TensorboardX*
- *Support hybridize to depoly*
- *Fast, 45ms/per_image when we resize max_side to 784*
Thanks for the author's (@whai362) great work!
## Requirements
- Python 2.7
- mxnet1.4.0
- pyclipper
- Polygon2
- OpenCV 4+ (for c++ version pse)
- TensorboardX
## Introduction
To reimplement PSENet by Gluon, here are some problem that I occur.
#### Diceloss about kernels isn't convergence.
- First, I doubt the label about kernel is not correct. However, I verify them again so that they are absolute right.
- Second, I doubt the `mx.nd.split` cannot be backwarded. However the diceloss about score map by `split` is well. So it cannot be raise this problem.
- Here the network is based on resnet50, and the output of FPN is *input_size/4*,so there may not be any text instance in min_kernel_map. So I set the number of kernels to *3*
Maybe upsampling output to input_size is a good choice. I will try it in my spare time.
#### Evaluation
| Dataset | Recall | Precision | F1-score | Speed |
| ------------------ | ------ | --------- | -------- | -------------- |
| ICPR(max_side=784) | 0.56 | 0.67 | 0.61 | **45**ms/image |
## Usage
#### Pretrained-models
- [gluoncv_model_zoo](https://gluon-cv.mxnet.io/model_zoo/classification.html):**resnet50_v1b**, you can replace it with others，the default path of pretrained-model in `~/.mxnet/`
Also you can download maskrcnn_coco from `gluoncv_model_zoo` to get a warm start.
#### Make
```
cd pse
make
```
Here I add `-Wl,-undefined,dynamic_lookup` to avoid some compile error, which is different from original PSENet.
#### Train
```
python scripts/train.py $data_path $ckpt
```
- `data_path`: path of dataset, which the prefix of image and annoation must be same, for example, a.jpg, a.txt
- `ckpt`: the filename of pretrained-mdel
#### Loss curve:
|  |  |  |  |
| :----------------------------------------------------------: | :----------------------------------------------------------: | :----------------------------------------------------------: | :----------------------------------------------------------: |
| Text loss | Kernel loss | All_loss | Pixel_accuracy |
#### Some Results
#### Inference
```
python eval.py $data_path $ckpt $output_dir $gpu_or_cpu
```
#### TODO:
- Upsamping to input_size
- Train on ICDAR and evaluate
### References
- [issue 15](https://github.com/whai362/PSENet/issues/15),
- [tensorflow_PSENET](https://github.com/liuheng92/tensorflow_PSENet)
- [issue10](https://github.com/whai362/PSENet/issues/10)
- [PSENet](https://github.com/whai362/PSENet)