Presentation | 2019-06-13 Accuracy Improvement of Depth Estimation from a Single Still Image Using Feature Pyramid Network Yudai Fukuda, Takuro Oki, Ryusuke Miyamoto, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Depth estimation from a single shot image have become accurate drastically after emergence of deepneural networks that show excellent results in several image recognition tasks. An existing scheme attempts to usenot only local features but global features by a feature pyramid composed of various scales of features obtainedfrom several convolution layers in order to improve the estimation accuracy. However, there is room for accuracyimprovement because generation of a feature pyramid in the scheme is not sophisticated. This paper tries to applythe Feature Pyramid Network for improve depth estimation in a single shot image, which was first adopted to objectdetection. Experimental results using the KITTI dataset showed that RMSE was improved by about 5% by theproposed scheme and accuracy became better than state-of-the-art schemes when the threshold was 1.25 3 .r |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | depth estimation / a single still image / pyramid feature network / deep learning |
Paper # | SIS2019-5 |
Date of Issue | 2019-06-06 (SIS) |
Conference Information | |
Committee | SIS / IPSJ-AVM / ITE-3DIT |
---|---|
Conference Date | 2019/6/13(2days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | Fukue Culture Center |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | Intelligent Multimedia Systems, Applied Enbedded Systems, Three-Dimensional Image Technology (3DIT), etc. |
Chair | Takayuki Nakachi(NTT) / Sei Naito(KDDI Research, Inc.) / Tsutomu Horikoshi(Shonan Institute of Technology) |
Vice Chair | Noriaki Suetake(Yamaguchi Univ.) / Tomoaki Kimura(Kanagawa Inst. of Tech.) |
Secretary | Noriaki Suetake(Tokyo Metropolitan Univ.) / Tomoaki Kimura(Kindai Univ.) / (NTT) / (Tokyo Univ. of Science) |
Assistant | Hideaki Misawa(National Inst. of Tech., Ube College) / Yukihiro Bandoh(NTT) |
Paper Information | |
Registration To | Technical Committee on Smart Info-Media Systems / Special Interest Group on Audio Visual and Multimedia Information Processing / Technical Group on Three-Dimensional Image Technology |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Accuracy Improvement of Depth Estimation from a Single Still Image Using Feature Pyramid Network |
Sub Title (in English) | |
Keyword(1) | depth estimation |
Keyword(2) | a single still image |
Keyword(3) | pyramid feature network |
Keyword(4) | deep learning |
1st Author's Name | Yudai Fukuda |
1st Author's Affiliation | Meiji University(Meiji Univ.) |
2nd Author's Name | Takuro Oki |
2nd Author's Affiliation | Meiji University(Meiji Univ.) |
3rd Author's Name | Ryusuke Miyamoto |
3rd Author's Affiliation | Meiji University(Meiji Univ.) |
Date | 2019-06-13 |
Paper # | SIS2019-5 |
Volume (vol) | vol.119 |
Number (no) | SIS-78 |
Page | pp.pp.23-28(SIS), |
#Pages | 6 |
Date of Issue | 2019-06-06 (SIS) |