KOYAMA So, a second-year master’s student in the Network Design Program at the Graduate School of Advanced Mathematical Sciences (FUKUYAMA Yoshikazu Laboratory), received the 2024 Institute of Electrical Engineers of Japan (IEEJ) Excellent Presentation Award at the IEE-Japan Industry Applications Society Conference (JIASC).
Organized by its technical committee, JIASC presents the Excellent Presentation Award to outstanding paper presentations by researchers under 35 that exemplify the qualities expected of promising young engineers.
KOYAMA’s presentation, “Application of Efficient GANs to Showcase Anomaly Detection: Comparison with Traditional Approaches,” explored techniques for identifying malfunctions in refrigerated and frozen showcases.
Awarded paper: KOYAMA So, FUKUYAMA Yoshikazu, MURAKAMI Kenya, SUZUKI Satoshi, IIZAKA Tatsuya: “Application of Efficient GANs to Showcase Anomaly Detection: Comparison with Traditional Approaches” IEEJ Joint Technical Meeting on Systems/Smart Facilities, ST-24 025, SMF-24 052, October 31, 2024
Refrigerated and frozen showcases are commonly used in supermarkets, convenience stores and cafeterias to store and display fresh and frozen foods as well as beverages. Although infrequent, malfunctions such as frost formation and refrigerant leakage can occur in these showcases, resulting in product disposal and lost sales opportunities. To mitigate these risks, accurate anomaly detection is essential.
This paper validated the effectiveness of the proposed Efficient Generative Adversarial Networks (GANs) for anomaly detection in showcases through the following procedures: First, the performance was compared with that of traditional approaches using the Area Under the Curve metric, which does not require specifying a threshold. Then, the simulated results were analyzed with Friedman’s test, followed by a post-hoc analysis using the Wilcoxon signed-rank test with adjustments applied through Holm’s method.
