Presentation | 2022-08-26 Object Recognition by Fingertip Gestures Utilizing Transfer Learning of YOLOv5 Tsukasa Kudo, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | In object recognition for videos and images, when the target area is small, the area is firstly extracted by object detection, and then object recognition is performed for that area. Various methods utilizing deep learning have been proposed to simultaneously perform object detection and recognition. However, when targeting various objects, there is a problem that it is necessary to prepare training data for all objects. To address this issue, in my previous study, I showed that a hand gesture trajectory can be recognized and then the target area can be extracted by utilizing a hand detection model using YOLOv5. In this study, I propose a method to improve the accuracy of the target area extraction. First, I show that the model can be trained efficiently by transfer learning to perform the gesture with a smaller fingertip. Next, I show that by utilizing the action of stopping the gesture, the gesture's interval indicating the target area and the feature points of the target can be extracted. Furthermore, experiments using books show that these methods can extract the books' target area as a tilt-corrected rectangle. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | YOLO / object detection / object recognition / gesture recognition / transfer learning |
Paper # | SWIM2022-12,SC2022-18 |
Date of Issue | 2022-08-19 (SWIM, SC) |
Conference Information | |
Committee | SWIM / SC |
---|---|
Conference Date | 2022/8/26(1days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Held online |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | Service Computing and Interprise Technology Toward Sciety 5.0 and etc |
Chair | Kenji Saotome(Hosei Univ.) / Kosaku Kimura(Fujitsu Lab.) |
Vice Chair | Akihiro Hayashi(Shizuoka Inst. of Science and Tech.) / Takao Nakaguchi(KCGI) |
Secretary | Akihiro Hayashi(Tokyo Univ. of Science) / Takao Nakaguchi(Osaka Sangyo Univ.) |
Assistant | Tsukasa Kudo(Shizuoka Inst. of Science and Tech.) / Kokichi Tsuji(Aichi Pref. Univ.) / Shigeru Hosono(Tokyo Univ. of Tech.) |
Paper Information | |
Registration To | Technical Committee on Software Interprise Modeling / Technical Committee on Service Computing |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Object Recognition by Fingertip Gestures Utilizing Transfer Learning of YOLOv5 |
Sub Title (in English) | |
Keyword(1) | YOLO |
Keyword(2) | object detection |
Keyword(3) | object recognition |
Keyword(4) | gesture recognition |
Keyword(5) | transfer learning |
1st Author's Name | Tsukasa Kudo |
1st Author's Affiliation | Shizuoka Institute of Science and Technology(SIST) |
Date | 2022-08-26 |
Paper # | SWIM2022-12,SC2022-18 |
Volume (vol) | vol.122 |
Number (no) | SWIM-167,SC-168 |
Page | pp.pp.9-15(SWIM), pp.9-15(SC), |
#Pages | 7 |
Date of Issue | 2022-08-19 (SWIM, SC) |