| Presentation | 2021-12-16 Real-world Anomaly Detection by Integrating Long-Term and Short-Term Classifications Yudai Watanabe, Makoto Okabe, Yasunori Harada, Yasuhiko Naoji, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | We propose a model to determine whether a video is normal or abnormal by using all the segments obtained from the video as input, and introducing a self-attention mechanism to analyze all the input segments and extract features that are important for determining normal/abnormal. The proposed method does not require MIL because it learns per video instead of per segment. During inference, the target video is divided into multiple short videos, and the proposed method is applied to each divided video to detect normal/abnormal per short time. We evaluated the frame-level detection accuracy of the proposed method on two benchmark datasets, and found that the proposed method can achieve the comparable accuracy as the state-of-the-art methods. In addition, the accuracy can be further improved by ensembling with other methods. We show that by averaging the results of the long-term and short-term detectors, we can achieve better detection accuracy than state-of-the-art methods. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Video anomaly detection / surveillance camera / weakly supervised learning / ensemble |
| Paper # | PRMU2021-27 |
| Date of Issue | 2021-12-09 (PRMU) |
| Conference Information | |
| Committee | PRMU |
|---|---|
| Conference Date | 2021/12/16(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Online |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Seiichi Uchida(Kyushu Univ.) |
| Vice Chair | Masakazu Iwamura(Osaka Pref. Univ.) / Mitsuru Anpai(Denso IT Lab.) |
| Secretary | Masakazu Iwamura(NTT) / Mitsuru Anpai(Tottori Univ.) |
| Assistant | Kouta Yamaguchi(CyberAgent) / Yusuke Matsui(Univ. of Tokyo) |
| Paper Information | |
| Registration To | Technical Committee on Pattern Recognition and Media Understanding |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Real-world Anomaly Detection by Integrating Long-Term and Short-Term Classifications |
| Sub Title (in English) | |
| Keyword(1) | Video anomaly detection |
| Keyword(2) | surveillance camera |
| Keyword(3) | weakly supervised learning |
| Keyword(4) | ensemble |
| 1st Author's Name | Yudai Watanabe |
| 1st Author's Affiliation | Shizuoka University(Shizuoka Univ.) |
| 2nd Author's Name | Makoto Okabe |
| 2nd Author's Affiliation | Shizuoka University(Shizuoka Univ.) |
| 3rd Author's Name | Yasunori Harada |
| 3rd Author's Affiliation | Chubu Electric Power Co., Inc.(Chubu Electric Power Co., Inc.) |
| 4th Author's Name | Yasuhiko Naoji |
| 4th Author's Affiliation | Chubu Electric Power Co., Inc.(Chubu Electric Power Co., Inc.) |
| Date | 2021-12-16 |
| Paper # | PRMU2021-27 |
| Volume (vol) | vol.121 |
| Number (no) | PRMU-304 |
| Page | pp.pp.19-24(PRMU), |
| #Pages | 6 |
| Date of Issue | 2021-12-09 (PRMU) |