A normalized cross-correlation convolutional neural network (CNN-NCC) for exemplar-based object detection.

Autor:	Wihardi, Yaya, Kristy, Winda M., Erlangga, Turnip, Arjon, Yulita, Intan N., Endroyono
Předmět:	CONVOLUTIONAL neural networks OBJECT recognition (Computer vision) VIDEO surveillance CROSS correlation ROBOTICS
Zdroj:	AIP Conference Proceedings; 2023, Vol. 2734 Issue 1, p1-7, 7p
Abstrakt:	Object detection in dynamic environments, such as CCTV surveillance systems and robotics, requires a special approach. New objects that have not been previously trained to the model are often captured by sensors, and it is very important to recognize them in the future. One approach that can be used to solve this problem is an exemplar-based approach. On this study, we try to build an exemplar-based object detector based on Convolutional Neural Network with Normalized Cross Correlation matching. Our proposed method did not need to re-train the model to identify unknown objects that had not been acknowledged by the model during the training stage. The results show that the model achieves 0.449 mAP on the COCO dataset. [ABSTRACT FROM AUTHOR]
Databáze:	Complementary Index
Externí odkaz:	Zobrazit plný text záznamu