| Presentation | 2021-03-29 A 3DCNN with Reduced Parameters Using Depthwise Separable Convolution Koki Ito, Hidehiro Nakano, Arata Miyauchi, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Convolutional Neural Networks (CNNs) have been used in various fields such as image and speech. In recent years, CNNs have been used not only for 2D images but also for 3D video images. However, these 3-Dimensional CNN (3DCNN) architectures are models that have evolved to compete for the highest accuracy in specific tasks, and the computational complexity and number of parameters have not been discussed so far. This fact has become an obstacle to the application of 3DCNNs. In this paper, we propose a 3DCNN architecture that can drastically reduce the number of parameters and still maintain the same recognition accuracy among networks that handle 3D information. In our experiments, we have succeeded in reducing the number of parameters by 94.6% in the task of human action recognition. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Deep Learning / Convolutional Neural Network / Human Action Recognition / Depthwise Separable Convolution |
| Paper # | CCS2020-27 |
| Date of Issue | 2021-03-22 (CCS) |
| Conference Information | |
| Committee | CCS |
|---|---|
| Conference Date | 2021/3/29(1days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Online |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | etc. |
| Chair | Shigeki Shiokawa(Kanagawa Inst. of Tech.) |
| Vice Chair | Tetsuya Asai(Hokkaido Univ.) / Megumi Akai(Hokkaido Univ.) |
| Secretary | Tetsuya Asai(Kanagawa Inst. of Tech.) / Megumi Akai(TDK) |
| Assistant | Hidehiro Nakano(Tokyo City Univ.) / Hiroyasu Ando(Tsukuba Univ.) / Takashi Matsubara(Kobe Univ.) / Kosuke Sanada(Mie Univ.) |
| Paper Information | |
| Registration To | Technical Committee on Complex Communication Sciences |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | A 3DCNN with Reduced Parameters Using Depthwise Separable Convolution |
| Sub Title (in English) | |
| Keyword(1) | Deep Learning |
| Keyword(2) | Convolutional Neural Network |
| Keyword(3) | Human Action Recognition |
| Keyword(4) | Depthwise Separable Convolution |
| 1st Author's Name | Koki Ito |
| 1st Author's Affiliation | Tokyo City University(Tokyo City Univ.) |
| 2nd Author's Name | Hidehiro Nakano |
| 2nd Author's Affiliation | Tokyo City University(Tokyo City Univ.) |
| 3rd Author's Name | Arata Miyauchi |
| 3rd Author's Affiliation | Tokyo City University(Tokyo City Univ.) |
| Date | 2021-03-29 |
| Paper # | CCS2020-27 |
| Volume (vol) | vol.120 |
| Number (no) | CCS-438 |
| Page | pp.pp.37-41(CCS), |
| #Pages | 5 |
| Date of Issue | 2021-03-22 (CCS) |