Presentation | 2022-09-29 Tuning properties to Second-order grating in Deep Convolutional Neural Network (DCNN) Jamal Shah, Yoshiyuki R. Shiraishi, Tomoyuki Naito, Takahisa M. Sanada, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Second-order visual feature, such as texture defined edge or contrast defined edge, is an important visual cue for object recognition. In the cortical visual system, neurons selective to second-order feature were found in area 18 and these neurons showed similar selectivity to luminance defined grating and the second-order gratings [1][2] Object recognition ability of Deep Convolutional Neural Network (DCNN) achieved the same level as humans. Since both cortical visual system and DCNN have hierarchical architecture, the DCNN might also process the 2nd-order feature in the middle layers. Here we analyzed response properties of internal units in VGG16 to luminance defined grating and the Second-order grating stimuli. VGG16 shows high neural predictability of area V4 [3]. By varying orientation and spatial frequency of luminance grating and envelope of the Second-order grating, we measured tuning characteristics of internal units in all layers. Many of units in middle layers (such as Block3) showed clear selectivity to envelope orientation and spatial frequency. Some of them showed similar tuning properties both in luminance defined grating and Second-order grating. This result suggests that VGG16 might follow similar manner for object representation as cortical visual system. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Receptive field / Visual cortex / Orientation tuning / Object recognition |
Paper # | NC2022-36 |
Date of Issue | 2022-09-22 (NC) |
Conference Information | |
Committee | NC / MBE |
---|---|
Conference Date | 2022/9/29(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Tohoku Univ. |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | Brain Architecture, NC, ME |
Chair | Hiroshi Yamakawa(Univ of Tokyo) / Junichi Hori(Niigata Univ.) |
Vice Chair | Hirokazu Tanaka(Tokyo City Univ.) / Hisashi Yoshida(Kinki Univ.) |
Secretary | Hirokazu Tanaka(NTT) / Hisashi Yoshida(NICT) |
Assistant | Yoshimasa Tawatsuji(Waseda Univ.) / Tomoki Kurikawa(KMU) / Emi Yuda(Tohoku Univ) / Miki Kaneko(Osaka Univ.) |
Paper Information | |
Registration To | Technical Committee on Neurocomputing / Technical Committee on ME and Bio Cybernetics |
---|---|
Language | ENG-JTITLE |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Tuning properties to Second-order grating in Deep Convolutional Neural Network (DCNN) |
Sub Title (in English) | |
Keyword(1) | Receptive field |
Keyword(2) | Visual cortex |
Keyword(3) | Orientation tuning |
Keyword(4) | Object recognition |
1st Author's Name | Jamal Shah |
1st Author's Affiliation | Iwate Prefectural University(Iwate Prefectural Univ.) |
2nd Author's Name | Yoshiyuki R. Shiraishi |
2nd Author's Affiliation | HobHop, Inc.(HobHop, Inc.) |
3rd Author's Name | Tomoyuki Naito |
3rd Author's Affiliation | Graduate School of Medicine, Osaka University(Osaka Univ.) |
4th Author's Name | Takahisa M. Sanada |
4th Author's Affiliation | Iwate Prefectural University(Iwate Prefectural Univ.) |
Date | 2022-09-29 |
Paper # | NC2022-36 |
Volume (vol) | vol.122 |
Number (no) | NC-195 |
Page | pp.pp.16-19(NC), |
#Pages | 4 |
Date of Issue | 2022-09-22 (NC) |