Presentation | 2021-01-20 Automatic Reading Of Industrial Meters using CNN with Generative Augmentation Sripanuskul Nuntida, Buayai Prawit, Xiaoyang Mao, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Although Convolutional Neural Network (CNN) showed high potential for automatic meter reading, it is facing various challenges. One is the need of a large quantity of training data while collecting the meter image dataset is an expensive task especially in the rotating meter where some digits take a long time to update. Another challenging issue is the existence of the transitional state between two consecutive numbers. To solve these problems, we propose a new automatic rotating meter reading framework featuring novel data augmentation techniques that can automatically generate annotated images of numbers including the transitional states between two consecutive numbers. By taking advantage of a generative neural network, the generated numbers resemble the local appearance of those in original meter images. Evaluation experiments confirm that our proposed generative data augmentation techniques improve the robustness of the recognition model and achieve outstanding results comparing to the previous work. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Automatic meter readingImage augmentationObject detectionAuto-encodeGenerative neural network |
Paper # | SeMI2020-48 |
Date of Issue | 2021-01-13 (SeMI) |
Conference Information | |
Committee | SeMI |
---|---|
Conference Date | 2021/1/20(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Online |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | |
Chair | Susumu Ishihara(Shizuoka Univ.) |
Vice Chair | Kazuya Monden(Hitachi) / Koji Yamamoto(Kyoto Univ.) |
Secretary | Kazuya Monden(Kyoto Univ.) / Koji Yamamoto(Cyber Univ.) |
Assistant | Yuki Katsumata(NTT DOCOMO) / Yu Nakayama(Tokyo Univ. of Agri. and Tech.) / Akira Uchiyama(Osaka Univ.) |
Paper Information | |
Registration To | Technical Committee on Sensor Network and Mobile Intelligence |
---|---|
Language | ENG |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Automatic Reading Of Industrial Meters using CNN with Generative Augmentation |
Sub Title (in English) | |
Keyword(1) | Automatic meter readingImage augmentationObject detectionAuto-encodeGenerative neural network |
1st Author's Name | Sripanuskul Nuntida |
1st Author's Affiliation | University of Yamanashi(Univ. of Yamanashi) |
2nd Author's Name | Buayai Prawit |
2nd Author's Affiliation | University of Yamanashi(Univ. of Yamanashi) |
3rd Author's Name | Xiaoyang Mao |
3rd Author's Affiliation | University of Yamanashi(Univ. of Yamanashi) |
Date | 2021-01-20 |
Paper # | SeMI2020-48 |
Volume (vol) | vol.120 |
Number (no) | SeMI-315 |
Page | pp.pp.27-32(SeMI), |
#Pages | 6 |
Date of Issue | 2021-01-13 (SeMI) |