Presentation | 2020-05-29 A method for analyze causes of deterioration of predict quality when Deep Learning is applied to instance segmentation Tomonori Kubota, Takanori Nakao, Masafumi Katoh, Eiji Yoshida, Hidenobu Miyoshi, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | In this paper, we propose a method to analyze the cause of deterioration of prediction accuracy in instance segmentation by deep learning. We have proposed a method to analyze the cause of deterioration of prediction accuracy in object recognition and object detection. This method is extended to instance segmentation (Mask Scoring R-CNN). This method extracts and visualizes the cause at the pixel grain size in the image (input image) in which the quality of the prediction result deteriorates. And, by applying the cause information of the pixel grain size extracted by this technique to the input image, it can be corrected to the image with improved prediction accuracy. That is, it can be shown that the cause information extracted by this method correctly represents the cause. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | deep learning / convolutional neural network / video analysis / segmentation / XAI |
Paper # | SIP2020-14,BioX2020-14,IE2020-14,MI2020-14 |
Date of Issue | 2020-05-21 (SIP, BioX, IE, MI) |
Conference Information | |
Committee | MI / IE / SIP / BioX / ITE-IST / ITE-ME |
---|---|
Conference Date | 2020/5/28(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | 会議ツールは未定 |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | Image and signal processing/analysis/AI technology, and their application |
Chair | Yoshiki Kawata(Tokushima Univ.) / Hideaki Kimata(NTT) / Naoyuki Aikawa(TUS) / Akira Otsuka(IISEC) / Shigetoshi Sugawa(Tohoku Univ.) / Arai Hiroyuki(Nippon Institute of Technology) |
Vice Chair | Takayuki Kitasaka(Aichi Inst. of Tech.) / Hidekata Hontani(Nagoya Inst. of Tech.) / Kazuya Kodama(NII) / Keita Takahashi(Nagoya Univ.) / Kazunori Hayashi(Osaka City Univ) / Yukihiro Bandou(NTT) / Tetsushi Ohki(Shizuoka Univ.) / Takahiro Aoki(Fujitsu Labs.) / Takayuki Hamamoto(Tokyo Univ. of Science) |
Secretary | Takayuki Kitasaka(Yamaguchi Univ.) / Hidekata Hontani(Univ. of Hyogo) / Kazuya Kodama(NTT) / Keita Takahashi(NHK) / Kazunori Hayashi(Hiroshima Univ.) / Yukihiro Bandou(Hosei Univ.) / Tetsushi Ohki(Univ. of Electro-Comm.) / Takahiro Aoki(SECOM) / Takayuki Hamamoto(Saitama Univ.) / (Panasonic) |
Assistant | Hotaka Takizawa(Tsukuba Univ.) / Yoshito Otake(NAIST) / Kyohei Unno(KDDI Research) / Norishige Fukushima(Nagoya Inst. of Tech.) / Kenjiro Sugimoto(Waseda Univ.) / Daishi Watabe(Saitama Inst. of Tech.) / Ryota Horie(Shibaura Inst. of Tech.) |
Paper Information | |
Registration To | Technical Committee on Medical Imaging / Technical Committee on Image Engineering / Technical Committee on Signal Processing / Technical Committee on Biometrics / Technical Group on Information Sensing Technologies / Technical Group on Media Engineering |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | A method for analyze causes of deterioration of predict quality when Deep Learning is applied to instance segmentation |
Sub Title (in English) | |
Keyword(1) | deep learning |
Keyword(2) | convolutional neural network |
Keyword(3) | video analysis |
Keyword(4) | segmentation |
Keyword(5) | XAI |
1st Author's Name | Tomonori Kubota |
1st Author's Affiliation | Fujitsu Laboratories LTD.(Fujitsu Lab.) |
2nd Author's Name | Takanori Nakao |
2nd Author's Affiliation | Fujitsu Laboratories LTD.(Fujitsu Lab.) |
3rd Author's Name | Masafumi Katoh |
3rd Author's Affiliation | Fujitsu Laboratories LTD.(Fujitsu Lab.) |
4th Author's Name | Eiji Yoshida |
4th Author's Affiliation | Fujitsu Laboratories LTD.(Fujitsu Lab.) |
5th Author's Name | Hidenobu Miyoshi |
5th Author's Affiliation | Fujitsu Laboratories LTD.(Fujitsu Lab.) |
Date | 2020-05-29 |
Paper # | SIP2020-14,BioX2020-14,IE2020-14,MI2020-14 |
Volume (vol) | vol.120 |
Number (no) | SIP-38,BioX-37,IE-39,MI-40 |
Page | pp.pp.67-72(SIP), pp.67-72(BioX), pp.67-72(IE), pp.67-72(MI), |
#Pages | 6 |
Date of Issue | 2020-05-21 (SIP, BioX, IE, MI) |