Floating Waste Discovery by Request via Object-Centric Learning

作     者：Bingfei Fu 

作者机构：School of Computer ScienceFudan UniversityShanghai200438China 

出 版 物：《Computers, Materials & Continua》 (计算机、材料和连续体（英文）)

年 卷 期：2024年第80卷第7期

页      面：1407-1424页

核心收录：

学科分类：08[工学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：Fudan University 

主　　题：Unsupervised object discovery object-centric learning pseudo data generation real-world object discovery by request 

摘      要：Discovering floating wastes,especially bottles on water,is a crucial research problem in environmental ***,real-world applications often face challenges such as interference from irrelevant objects and the high cost associated with data ***,devising algorithms capable of accurately localizing specific objects within a scene in scenarios where annotated data is limited remains a formidable *** solve this problem,this paper proposes an object discovery by request problem setting and a corresponding algorithmic *** proposed problem setting aims to identify specified objects in scenes,and the associated algorithmic framework comprises pseudo data generation and object discovery by request ***-data generation generates images resembling natural scenes through various data augmentation rules,using a small number of object samples and scene *** network structure of object discovery by request utilizes the pre-trained Vision Transformer(ViT)model as the backbone,employs object-centric methods to learn the latent representations of foreground objects,and applies patch-level reconstruction constraints to the *** the validation phase,we use the generated pseudo datasets as training sets and evaluate the performance of our model on the original test *** have proved that our method achieves state-of-the-art performance on Unmanned Aerial Vehicles-Bottle Detection(UAV-BD)dataset and self-constructed dataset Bottle,especially in multi-object scenarios.