https://github.com/wwwwwydev/pen-stroke-segmentation

基于深度学习中全卷积网络的字体笔划分割（汉字笔画分割）解决方案 A Solution for Font Stroke Segmentation (Stroke Segmentation) Based on Fully Convolutional Networks in Deep Learning
https://github.com/wwwwwydev/pen-stroke-segmentation

deep-learning deep-neural-networks pen-strokes python pytorch

Last synced: 30 days ago
JSON representation

基于深度学习中全卷积网络的字体笔划分割（汉字笔画分割）解决方案 A Solution for Font Stroke Segmentation (Stroke Segmentation) Based on Fully Convolutional Networks in Deep Learning

• Host: GitHub
• URL: https://github.com/wwwwwydev/pen-stroke-segmentation
• Owner: WwwwwyDev
• License: mit
• Created: 2024-04-03T08:18:29.000Z (10 months ago)
• Default Branch: master
• Last Pushed: 2024-04-24T08:23:29.000Z (10 months ago)
• Last Synced: 2024-11-12T13:17:08.084Z (3 months ago)
• Topics: deep-learning, deep-neural-networks, pen-strokes, python, pytorch
• Language: Python
• Homepage:
• Size: 347 KB
• Stars: 2
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        


  



#### 介绍

针对当前汉字笔画提取存在识别精度低和笔画分割不清等问题, 提出了一种辅助字体设计的汉字字体笔画分割模型. 该模型可对汉字图像进行有效的笔画分割。方法引入跳跃结构加强了笔画结构信息损失; 针对字体语义信息丢失的问题, 将模型中最大池化改为空洞卷积, 利用空洞卷积减少下采样的信息损失, 提升分割精度.

#### 论文地址

[论文地址](https://kns.cnki.net/kcms/detail/detail.aspx?dbcode=CJFD&dbname=CJFDAUTO&filename=DLMY202205015&uniplatform=NZKPT&v=Iu5O7u8Ge54PiIWXii5d02_IbkDVWZxMQRpGIp3SHVmXw2wJesFn2Nt_2fo48rpD)

#### 数据集

[百度网盘](https://pan.baidu.com/s/1ajN7HQkPbbPcmAuKJJXCNA?pwd=f7tU)  提取码: f7tU 

#### 预训练模型

[百度网盘](https://pan.baidu.com/s/1YZkR2ezE27ZshPHmsGtE5A)  提取码：fgpk

#### 网络结构

![输入图片说明](img/%E6%88%AA%E5%B1%8F2022-06-15%20%E4%B8%8B%E5%8D%888.06.43.png)

#### 分割效果

![输入图片说明](img/%E5%9B%BE11.jpg)