id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_tkkogouk5ffjfmynlbvfpjv6qu	Zhihui Wang	Multi-attention based cross-domain beauty product image retrieval	2020	3	.pdf	application/pdf	1948	218	62	Multi-attention based cross-domain beauty product image retrieval. held by ACM MultiMedia 2018 [1] for beauty product image retrieval task, and the Perfect-500K an end-to-end multi-attention classification network MANet, accounting for the basic feature extraction. Besides, a saliency-based regional maximum activation of convolutions (SR-MAC) module for feature representation is proposed to reduce the effect of background regions unrelated to related to the salient region; it is capable of objectively aggregating multiple local features and making the feature representation of beauty product images more discriminative. the product objects and the text regions in the images; (2) a robust local feature aggregation method mechanism; (3) a well-labeled beauty product image dataset is built, and the network is trained on images retrieval is proposed, focusing on the features of saliency regions and text regions in the product images, an SR-MAC feature representation module is proposed. beauty product dataset, Perfect-30K, with 44 categories for training our proposed MANet is constructed.	./cache/work_tkkogouk5ffjfmynlbvfpjv6qu.pdf	./txt/work_tkkogouk5ffjfmynlbvfpjv6qu.txt