| Presentation | 2019-05-10 Spatial-Separable Convolution: Low memory CNN for FPGA Akira Jinguji, Masayuki Shimoda, Hiroki Nakahara, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Object detection and image recognition using a Convolutional Neural Network (CNN) are used in em- bedded systems, which require reasonable price and power performance. Since CNN has high accuracy and large computation, real-time processing cannot be realized in CPU, and power consumption is too large in GPU. The CNN realization of the FPGA is low power consumption, however large on-chip memory is required and expensive. Typically, feature-map size in layers is large. This is a bottleneck in FPGA memory resource restrictions. We propose Feature-Map Separable Convolution, which makes an inference with divided feature-map. The feature-map size becomes smaller when an input image size becomes smaller. Thus, the buffer memory can be reduced. From experiments, we accomplished that the accuracy does not decrease so much with reducing buffer memory by 85%. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | FPGA / CNN |
| Paper # | RECONF2019-16 |
| Date of Issue | 2019-05-02 (RECONF) |
| Conference Information | |
| Committee | RECONF |
|---|---|
| Conference Date | 2019/5/9(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Tokyo Tech Front |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | Reconfigurable system, etc. |
| Chair | Masato Motomura(Tokyo Tech.) |
| Vice Chair | Yuichiro Shibata(Nagasaki Univ.) / Kentaro Sano(RIKEN) |
| Secretary | Yuichiro Shibata(Hiroshima City Univ.) / Kentaro Sano(e-trees.Japan) |
| Assistant | Yuuki Kobayashi(NEC) / Hiroki Nakahara(Tokyo Inst. of Tech.) |
| Paper Information | |
| Registration To | Technical Committee on Reconfigurable Systems |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Spatial-Separable Convolution: Low memory CNN for FPGA |
| Sub Title (in English) | |
| Keyword(1) | FPGA |
| Keyword(2) | CNN |
| 1st Author's Name | Akira Jinguji |
| 1st Author's Affiliation | Tokyo Institute of Tech(titech) |
| 2nd Author's Name | Masayuki Shimoda |
| 2nd Author's Affiliation | Tokyo Institute of Tech(titech) |
| 3rd Author's Name | Hiroki Nakahara |
| 3rd Author's Affiliation | Tokyo Institute of Tech(titech) |
| Date | 2019-05-10 |
| Paper # | RECONF2019-16 |
| Volume (vol) | vol.119 |
| Number (no) | RECONF-18 |
| Page | pp.pp.85-90(RECONF), |
| #Pages | 6 |
| Date of Issue | 2019-05-02 (RECONF) |