Home

Awesome

Shape Robust Text Detection with Progressive Scale Expansion Network

A reimplement of PSENet with mxnet-gluon. Just train on ICPR.

Thanks for the author's (@whai362) great work!

Requirements

Introduction

To reimplement PSENet by Gluon, here are some problem that I occur.

Diceloss about kernels isn't convergence.

Maybe upsampling output to input_size is a good choice. I will try it in my spare time.

Evaluation

DatasetRecallPrecisionF1-scoreSpeed
ICPR(max_side=784)0.560.670.6145ms/image

Usage

Pretrained-models

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

Loss curve:

image-20190614182216647image-20190614182249280image-20190614182313296image-20190614182326647
Text lossKernel lossAll_lossPixel_accuracy

Some Results

fusion_TB1vcxDLXXXXXb1XFXXunYpLFXX

Inference

python eval.py $data_path $ckpt $output_dir $gpu_or_cpu

TODO:

References