Synthetic Map Generation to Provide Unlimited Training Data for Historical Map Text Detection. Li, Z., Guan, R., Yu, Q., Chiang, Y., & Knoblock, C. A. In Proceedings of the 4th ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery, of GEOAI '21, pages 17–26, New York, NY, USA, 2021. Association for Computing Machinery. ![Synthetic Map Generation to Provide Unlimited Training Data for Historical Map Text Detection [link]](https://bibbase.org/img/filetypes/link.svg "link")
Paper ![Synthetic Map Generation to Provide Unlimited Training Data for Historical Map Text Detection [pdf]](https://bibbase.org/img/filetypes/pdf.svg "pdf")
Slides doi abstract bibtex 6 downloads Many historical map sheets are publicly available for studies that require long-term historical geographic data. The cartographic design of these maps includes a combination of map symbols and text labels. Automatically reading text labels from map images could greatly speed up the map interpretation and helps generate rich metadata describing the map content. Many text detection algorithms have been proposed to locate text regions in map images automatically, but most of the algorithms are trained on out-of-domain datasets (e.g., scenic images). Training data determines the quality of machine learning models, and manually annotating text regions in map images is labor-extensive and time-consuming. On the other hand, existing geographic data sources, such as Open-StreetMap (OSM), contain machine-readable map layers, which allow us to separate out the text layer and obtain text label annotations easily. However, the cartographic styles between OSM map tiles and historical maps are significantly different. This paper proposes a method to automatically generate an unlimited amount of annotated historical map images for training text detection models. We use a style transfer model to convert contemporary map images into historical style and place text labels upon them. We show that the state-of-the-art text detection models (e.g., PSENet) can benefit from the synthetic historical maps and achieve significant improvement for historical map text detection.
@inproceedings{10.1145/3486635.3491070,
author = {Li, Zekun and Guan, Runyu and Yu, Qianmu and Chiang, Yao-Yi and Knoblock, Craig A.},
title = {Synthetic Map Generation to Provide Unlimited Training Data for Historical Map Text Detection},
year = {2021},
isbn = {9781450391207},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3486635.3491070},
doi = {10.1145/3486635.3491070},
abstract = {Many historical map sheets are publicly available for studies that require long-term historical geographic data. The cartographic design of these maps includes a combination of map symbols and text labels. Automatically reading text labels from map images could greatly speed up the map interpretation and helps generate rich metadata describing the map content. Many text detection algorithms have been proposed to locate text regions in map images automatically, but most of the algorithms are trained on out-of-domain datasets (e.g., scenic images). Training data determines the quality of machine learning models, and manually annotating text regions in map images is labor-extensive and time-consuming. On the other hand, existing geographic data sources, such as Open-StreetMap (OSM), contain machine-readable map layers, which allow us to separate out the text layer and obtain text label annotations easily. However, the cartographic styles between OSM map tiles and historical maps are significantly different. This paper proposes a method to automatically generate an unlimited amount of annotated historical map images for training text detection models. We use a style transfer model to convert contemporary map images into historical style and place text labels upon them. We show that the state-of-the-art text detection models (e.g., PSENet) can benefit from the synthetic historical maps and achieve significant improvement for historical map text detection.},
booktitle = {Proceedings of the 4th ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery},
pages = {17–26},
numpages = {10},
keywords = {historical maps, text detection, datasets, synthetic data generation},
location = {Beijing, China},
series = {GEOAI '21},
URLslides = "http://usc-isi-i2.github.io/slides/li-sigspatial21-slides.pdf"
}
Downloads: 6
{"_id":"G6KvZET6MN7LhDsqE","bibbaseid":"li-guan-yu-chiang-knoblock-syntheticmapgenerationtoprovideunlimitedtrainingdataforhistoricalmaptextdetection-2021","author_short":["Li, Z.","Guan, R.","Yu, Q.","Chiang, Y.","Knoblock, C. A."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","author":[{"propositions":[],"lastnames":["Li"],"firstnames":["Zekun"],"suffixes":[]},{"propositions":[],"lastnames":["Guan"],"firstnames":["Runyu"],"suffixes":[]},{"propositions":[],"lastnames":["Yu"],"firstnames":["Qianmu"],"suffixes":[]},{"propositions":[],"lastnames":["Chiang"],"firstnames":["Yao-Yi"],"suffixes":[]},{"propositions":[],"lastnames":["Knoblock"],"firstnames":["Craig","A."],"suffixes":[]}],"title":"Synthetic Map Generation to Provide Unlimited Training Data for Historical Map Text Detection","year":"2021","isbn":"9781450391207","publisher":"Association for Computing Machinery","address":"New York, NY, USA","url":"https://doi.org/10.1145/3486635.3491070","doi":"10.1145/3486635.3491070","abstract":"Many historical map sheets are publicly available for studies that require long-term historical geographic data. The cartographic design of these maps includes a combination of map symbols and text labels. Automatically reading text labels from map images could greatly speed up the map interpretation and helps generate rich metadata describing the map content. Many text detection algorithms have been proposed to locate text regions in map images automatically, but most of the algorithms are trained on out-of-domain datasets (e.g., scenic images). Training data determines the quality of machine learning models, and manually annotating text regions in map images is labor-extensive and time-consuming. On the other hand, existing geographic data sources, such as Open-StreetMap (OSM), contain machine-readable map layers, which allow us to separate out the text layer and obtain text label annotations easily. However, the cartographic styles between OSM map tiles and historical maps are significantly different. This paper proposes a method to automatically generate an unlimited amount of annotated historical map images for training text detection models. We use a style transfer model to convert contemporary map images into historical style and place text labels upon them. We show that the state-of-the-art text detection models (e.g., PSENet) can benefit from the synthetic historical maps and achieve significant improvement for historical map text detection.","booktitle":"Proceedings of the 4th ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery","pages":"17–26","numpages":"10","keywords":"historical maps, text detection, datasets, synthetic data generation","location":"Beijing, China","series":"GEOAI '21","urlslides":"http://usc-isi-i2.github.io/slides/li-sigspatial21-slides.pdf","bibtex":"@inproceedings{10.1145/3486635.3491070,\nauthor = {Li, Zekun and Guan, Runyu and Yu, Qianmu and Chiang, Yao-Yi and Knoblock, Craig A.},\ntitle = {Synthetic Map Generation to Provide Unlimited Training Data for Historical Map Text Detection},\nyear = {2021},\nisbn = {9781450391207},\npublisher = {Association for Computing Machinery},\naddress = {New York, NY, USA},\nurl = {https://doi.org/10.1145/3486635.3491070},\ndoi = {10.1145/3486635.3491070},\nabstract = {Many historical map sheets are publicly available for studies that require long-term historical geographic data. The cartographic design of these maps includes a combination of map symbols and text labels. Automatically reading text labels from map images could greatly speed up the map interpretation and helps generate rich metadata describing the map content. Many text detection algorithms have been proposed to locate text regions in map images automatically, but most of the algorithms are trained on out-of-domain datasets (e.g., scenic images). Training data determines the quality of machine learning models, and manually annotating text regions in map images is labor-extensive and time-consuming. On the other hand, existing geographic data sources, such as Open-StreetMap (OSM), contain machine-readable map layers, which allow us to separate out the text layer and obtain text label annotations easily. However, the cartographic styles between OSM map tiles and historical maps are significantly different. This paper proposes a method to automatically generate an unlimited amount of annotated historical map images for training text detection models. We use a style transfer model to convert contemporary map images into historical style and place text labels upon them. We show that the state-of-the-art text detection models (e.g., PSENet) can benefit from the synthetic historical maps and achieve significant improvement for historical map text detection.},\nbooktitle = {Proceedings of the 4th ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery},\npages = {17–26},\nnumpages = {10},\nkeywords = {historical maps, text detection, datasets, synthetic data generation},\nlocation = {Beijing, China},\nseries = {GEOAI '21},\nURLslides = \"http://usc-isi-i2.github.io/slides/li-sigspatial21-slides.pdf\"\n}\n\n","author_short":["Li, Z.","Guan, R.","Yu, Q.","Chiang, Y.","Knoblock, C. A."],"key":"10.1145/3486635.3491070","id":"10.1145/3486635.3491070","bibbaseid":"li-guan-yu-chiang-knoblock-syntheticmapgenerationtoprovideunlimitedtrainingdataforhistoricalmaptextdetection-2021","role":"author","urls":{"Paper":"https://doi.org/10.1145/3486635.3491070","Slides":"http://usc-isi-i2.github.io/slides/li-sigspatial21-slides.pdf"},"keyword":["historical maps","text detection","datasets","synthetic data generation"],"metadata":{"authorlinks":{}},"downloads":6},"bibtype":"inproceedings","biburl":"https://bibbase.org/network/files/3E8vjZ9tpMwaLo8Hx","dataSources":["ZRsZjZ5CrudyLys96","Kv8cz2f6c8JXfAMgD","kKMapWSMFgtAAqeuL","HAFwXuLZf7sqJvp2S","8u7qceCaxL8Gt5PoF","otmHudCn8d9FgQCrS","3sPtWLmmdPRfH69LS","TXXdv8JoYH4GANs8K","9XM4dtiSCrrfrXGGq","c3uZgDfvZw2qtJCbJ","LYRyuex4y2kvvDWRm","n4SD4EqQ5R79AWNE6"],"keywords":["historical maps","text detection","datasets","synthetic data generation"],"search_terms":["synthetic","map","generation","provide","unlimited","training","data","historical","map","text","detection","li","guan","yu","chiang","knoblock"],"title":"Synthetic Map Generation to Provide Unlimited Training Data for Historical Map Text Detection","year":2021,"downloads":6}