Traffic Signs: Berlin, 2020

  1. Introduction The main idea of ​​the project is to obtain traffic sign locations by analyzing videos using a combination of artificial intelligence and image recognition methods. Each video file also includes a geolocation file that has the same name as the video file and contains latitude, longitude, and timestamp attributes from the beginning of the video. A total of 3350 videos with a total range of 1040 km are used in the area of ​​the Berlin S-Bahn ring. The result file contains longitude and latitude (WGS84, EPSG:4326) of traffic sign locations and their types in 43 categories. 2. Data sets To train AI networks, two publicly available data sets are used: for traffic sign recognition “German Traffic Sign Detection Benchmark Dataset[1]” and for traffic sign classification “German Traffic Sign Recognition Benchmark Dataset[1 ]". You can find more information here: Detection Dataset, Classification Dataset 3. Methodology and models The TensorFlow[2] framework is used to analyze videos. An object detection[3] model for traffic sign recognition is trained using the transfer learning method[4]. To improve the accuracy of traffic sign classification, a custom image classification[5] model for categorizing traffic sign types is trained. The output of the traffic sign recognition model is used as input of the traffic sign classification model. 4. Source [1] Houben, S., Stallkamp, ​​J., Salmen, J., Schlipsing, M. and Igel, C. (2013). "Detection of traffic signs in real-world images: the German traffic sign detection benchmark", in Proceedings of the International Joint Conference on Neural Networks.10.1109/IJCNN.2013.6706807 [2] Martín A., Paul B ., Jianmin C, Zhifeng C, Andy D, Jeffrey D, .... Xiaoqiang Zheng. (2016). "TensorFlow: a system for large-scale machine learning", in Proceedings of the 12th USENIX conference on Operating Systems Design and Implementation(OSDI'16). USENIX Association, USA, 265–283 [3] Girshick R, Donahue J, Darrell T, Malik J (2014). "Rich feature hierarchies for accurate object detection and semantic segmentation", 2014 IEEE Conference on Computer Vision and Pattern Recognition. doi:10.1109/cvpr.2014.81 [4] Tan C, Sun F, Kong T, Zhang W, Yang C and Liu C (2018) “A survey on deep transfer learning”, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 11141 LNCS, pp. 270–279. doi: 10.1007/978-3-030-01424-7_27. [5] Sultana, F., Sufian, A. and Dutta, P. (2018). “Advancements in image classification using convolutional neural network”, in Proceedings - 2018 4th IEEE International Conference on Research in Computational Intelligence and Communication Networks, ICRCICN 2018, pp. 122–129. doi: 10.1109/ICRCCICN.2018.8718718.

Подаци и Ресурси

Додатне информације

Поље Вредност
Последње измене јун 17, 2023, 02:03 (UTC)
Креирано јун 17, 2023, 02:03 (UTC)
End of temporal extent 2020-09-30
Frequency http://publications.europa.eu/resource/authority/frequency/NEVER
GUID https://mcloud.de/web/guest/suche/-/results/detail/722EDEC3-38BA-4FE2-B087-18C0434CA34E
Identifier 722EDEC3-38BA-4FE2-B087-18C0434CA34E
Issued 2021-01-14T16:12:01+00:00
Modified 2023-03-06T11:04:53+00:00
Publisher URI https://mcloud.de/web/guest/suche/-/results/detail/722EDEC3-38BA-4FE2-B087-18C0434CA34E#publisher
Publisher name AIPARK GmbH
Start of temporal extent 2020-02-29
Theme ["http://publications.europa.eu/resource/authority/data-theme/TRAN"]
URI https://mcloud.de/web/guest/suche/-/results/detail/722EDEC3-38BA-4FE2-B087-18C0434CA34E
access_rights Der Datensatz wird für experimentelle Zwecke erstellt. Unsere Firma ist nicht verantwortlich für das Ergebnis möglicher Datenfehler.
contributorID ["http://dcat-ap.de/def/contributors/mCloud"]
harvest_object_id 83e8e145-6324-4179-bad4-5a2e882bd284
harvest_source_id 1da0a2ca-d24c-4f40-a127-12319109776d
harvest_source_title Germany - Open Data Portal
metadata_harvested_portal http://mcloud.de/
originator_contacttype Organization
originator_name Bliq GmbH, 2020
spatial {"type": "Polygon", "coordinates": [[[6.098034, 47.23942], [15.578555, 47.23942], [15.578555, 55.054344], [6.098034, 55.054344], [6.098034, 47.23942]]]}
spatial_text Bundesrepublik Deutschland