Difference between revisions of "Publications/chen.21.icdar"
From LRDE
(2 intermediate revisions by the same user not shown) | |||
Line 1: | Line 1: | ||
{{Publication |
{{Publication |
||
| published = true |
| published = true |
||
− | | date = |
+ | | date = 2021-05-17 |
| title = Vectorization of Historical Maps Using Deep Edge Filtering and Closed Shape Extraction |
| title = Vectorization of Historical Maps Using Deep Edge Filtering and Closed Shape Extraction |
||
| authors = Yizi Chen, Edwin Carlinet, Joseph Chazalon, Clément Mallet, Bertrand Duménieu, Julien Perret |
| authors = Yizi Chen, Edwin Carlinet, Joseph Chazalon, Clément Mallet, Bertrand Duménieu, Julien Perret |
||
| booktitle = Proceedings of the 16th International Conference on Document Analysis and Recognition (ICDAR'21) |
| booktitle = Proceedings of the 16th International Conference on Document Analysis and Recognition (ICDAR'21) |
||
− | | pages = |
+ | | pages = 510 to 525 |
+ | | series = Lecture Notes in Computer Science |
||
+ | | publisher = Springer, Cham |
||
+ | | volume = 12824 |
||
| address = Lausanne, Switzerland |
| address = Lausanne, Switzerland |
||
− | | abstract = Maps have been a unique source of knowledge for centuries. |
+ | | abstract = Maps have been a unique source of knowledge for centuries. Such historical documents provide invaluable information for analyzing the complex spatial transformation of landscapes over important time frames. This is particularly true for urban areas that encompass multiple interleaved research domains (social sciences, economy, etc.). The large amount and significant diversity of map sources call for automatic image processing techniques in order to extract the relevant objects under a vectorial shape. The complexity of maps (text, noise, digitization artifactsetc.) has hindered the capacity of proposing a versatile and efficient raster-to-vector approaches for decades. We propose a learnable, reproducible, and reusable solution for the automatic transformation of raster maps into vector objects (building blocks, streets, rivers). It is built upon the complementary strength of mathematical morphology and convolutional neural networks through efficient edge filtering. Evenmore, we modify ConnNet and combine with deep edge filtering architecture to make use of pixel connectivity information and built an end-to-end system without requiring any post-processing techniques. In this paper, we focus on the comprehensive benchmark on various architectures on multiple datasets coupled with a novel vectorization step. Our experimental results on a new public dataset using COCO Panoptic metric exhibit very encouraging results confirmed by a qualitative analysis of the success and failure cases of our approach. Codedataset, results and extra illustrations are freely available at https://github.com/soduco/ICDAR-2021-Vectorization. |
| lrdepaper = http://www.lrde.epita.fr/dload/papers/chen.21.icdar.pdf |
| lrdepaper = http://www.lrde.epita.fr/dload/papers/chen.21.icdar.pdf |
||
| lrdeprojects = Olena |
| lrdeprojects = Olena |
||
| lrdekeywords = Image |
| lrdekeywords = Image |
||
− | | lrdenewsdate = |
+ | | lrdenewsdate = 2021-05-17 |
− | | note = To appear |
||
| type = inproceedings |
| type = inproceedings |
||
| id = chen.21.icdar |
| id = chen.21.icdar |
||
+ | | identifier = doi:10.1007/978-3-030-86337-1_34 |
||
| bibtex = |
| bibtex = |
||
@InProceedings<nowiki>{</nowiki> chen.21.icdar, |
@InProceedings<nowiki>{</nowiki> chen.21.icdar, |
||
title = <nowiki>{</nowiki>Vectorization of Historical Maps Using Deep Edge Filtering |
title = <nowiki>{</nowiki>Vectorization of Historical Maps Using Deep Edge Filtering |
||
and Closed Shape Extraction<nowiki>}</nowiki>, |
and Closed Shape Extraction<nowiki>}</nowiki>, |
||
− | author = <nowiki>{</nowiki> |
+ | author = <nowiki>{</nowiki>Yizi Chen and Edwin Carlinet and Joseph Chazalon and |
Cl\'ement Mallet and Bertrand Dum\'enieu and Julien Perret<nowiki>}</nowiki>, |
Cl\'ement Mallet and Bertrand Dum\'enieu and Julien Perret<nowiki>}</nowiki>, |
||
booktitle = <nowiki>{</nowiki>Proceedings of the 16th International Conference on |
booktitle = <nowiki>{</nowiki>Proceedings of the 16th International Conference on |
||
Line 25: | Line 28: | ||
year = <nowiki>{</nowiki>2021<nowiki>}</nowiki>, |
year = <nowiki>{</nowiki>2021<nowiki>}</nowiki>, |
||
month = sep, |
month = sep, |
||
− | pages = <nowiki>{</nowiki><nowiki>}</nowiki>, |
+ | pages = <nowiki>{</nowiki>510--525<nowiki>}</nowiki>, |
+ | series = <nowiki>{</nowiki>Lecture Notes in Computer Science<nowiki>}</nowiki>, |
||
+ | publisher = <nowiki>{</nowiki>Springer, Cham<nowiki>}</nowiki>, |
||
+ | volume = <nowiki>{</nowiki>12824<nowiki>}</nowiki>, |
||
address = <nowiki>{</nowiki>Lausanne, Switzerland<nowiki>}</nowiki>, |
address = <nowiki>{</nowiki>Lausanne, Switzerland<nowiki>}</nowiki>, |
||
− | abstract = <nowiki>{</nowiki> |
+ | abstract = <nowiki>{</nowiki>Maps have been a unique source of knowledge for centuries. |
− | + | Such historical documents provide invaluable information |
|
− | + | for analyzing the complex spatial transformation of |
|
− | + | landscapes over important time frames. This is particularly |
|
− | + | true for urban areas that encompass multiple interleaved |
|
− | + | research domains (social sciences, economy, etc.). The |
|
− | + | large amount and significant diversity of map sources call |
|
− | + | for automatic image processing techniques in order to |
|
− | + | extract the relevant objects under a vectorial shape. The |
|
− | + | complexity of maps (text, noise, digitization artifacts, |
|
− | + | etc.) has hindered the capacity of proposing a versatile |
|
− | + | and efficient raster-to-vector approaches for decades. We |
|
− | + | propose a learnable, reproducible, and reusable solution |
|
⚫ | |||
− | reproducible, and reusable solution for the automatic |
||
⚫ | |||
⚫ | |||
⚫ | |||
⚫ | |||
⚫ | |||
⚫ | |||
⚫ | |||
filtering. Evenmore, we modify ConnNet and combine with |
filtering. Evenmore, we modify ConnNet and combine with |
||
deep edge filtering architecture to make use of pixel |
deep edge filtering architecture to make use of pixel |
||
Line 58: | Line 63: | ||
available at |
available at |
||
\url<nowiki>{</nowiki>https://github.com/soduco/ICDAR-2021-Vectorization<nowiki>}</nowiki>. <nowiki>}</nowiki>, |
\url<nowiki>{</nowiki>https://github.com/soduco/ICDAR-2021-Vectorization<nowiki>}</nowiki>. <nowiki>}</nowiki>, |
||
− | + | doi = <nowiki>{</nowiki>10.1007/978-3-030-86337-1_34<nowiki>}</nowiki> |
|
<nowiki>}</nowiki> |
<nowiki>}</nowiki> |
||
Latest revision as of 09:55, 8 September 2021
- Authors
- Yizi Chen, Edwin Carlinet, Joseph Chazalon, Clément Mallet, Bertrand Duménieu, Julien Perret
- Where
- Proceedings of the 16th International Conference on Document Analysis and Recognition (ICDAR'21)
- Place
- Lausanne, Switzerland
- Type
- inproceedings
- Publisher
- Springer, Cham
- Projects
- Olena
- Keywords
- Image
- Date
- 2021-05-17
Abstract
Maps have been a unique source of knowledge for centuries. Such historical documents provide invaluable information for analyzing the complex spatial transformation of landscapes over important time frames. This is particularly true for urban areas that encompass multiple interleaved research domains (social sciences, economy, etc.). The large amount and significant diversity of map sources call for automatic image processing techniques in order to extract the relevant objects under a vectorial shape. The complexity of maps (text, noise, digitization artifactsetc.) has hindered the capacity of proposing a versatile and efficient raster-to-vector approaches for decades. We propose a learnable, reproducible, and reusable solution for the automatic transformation of raster maps into vector objects (building blocks, streets, rivers). It is built upon the complementary strength of mathematical morphology and convolutional neural networks through efficient edge filtering. Evenmore, we modify ConnNet and combine with deep edge filtering architecture to make use of pixel connectivity information and built an end-to-end system without requiring any post-processing techniques. In this paper, we focus on the comprehensive benchmark on various architectures on multiple datasets coupled with a novel vectorization step. Our experimental results on a new public dataset using COCO Panoptic metric exhibit very encouraging results confirmed by a qualitative analysis of the success and failure cases of our approach. Codedataset, results and extra illustrations are freely available at https://github.com/soduco/ICDAR-2021-Vectorization.
Documents
Bibtex (lrde.bib)
@InProceedings{ chen.21.icdar, title = {Vectorization of Historical Maps Using Deep Edge Filtering and Closed Shape Extraction}, author = {Yizi Chen and Edwin Carlinet and Joseph Chazalon and Cl\'ement Mallet and Bertrand Dum\'enieu and Julien Perret}, booktitle = {Proceedings of the 16th International Conference on Document Analysis and Recognition (ICDAR'21)}, year = {2021}, month = sep, pages = {510--525}, series = {Lecture Notes in Computer Science}, publisher = {Springer, Cham}, volume = {12824}, address = {Lausanne, Switzerland}, abstract = {Maps have been a unique source of knowledge for centuries. Such historical documents provide invaluable information for analyzing the complex spatial transformation of landscapes over important time frames. This is particularly true for urban areas that encompass multiple interleaved research domains (social sciences, economy, etc.). The large amount and significant diversity of map sources call for automatic image processing techniques in order to extract the relevant objects under a vectorial shape. The complexity of maps (text, noise, digitization artifacts, etc.) has hindered the capacity of proposing a versatile and efficient raster-to-vector approaches for decades. We propose a learnable, reproducible, and reusable solution for the automatic transformation of raster maps into vector objects (building blocks, streets, rivers). It is built upon the complementary strength of mathematical morphology and convolutional neural networks through efficient edge filtering. Evenmore, we modify ConnNet and combine with deep edge filtering architecture to make use of pixel connectivity information and built an end-to-end system without requiring any post-processing techniques. In this paper, we focus on the comprehensive benchmark on various architectures on multiple datasets coupled with a novel vectorization step. Our experimental results on a new public dataset using COCO Panoptic metric exhibit very encouraging results confirmed by a qualitative analysis of the success and failure cases of our approach. Code, dataset, results and extra illustrations are freely available at \url{https://github.com/soduco/ICDAR-2021-Vectorization}. }, doi = {10.1007/978-3-030-86337-1_34} }