Regional Maximum Activations of Convolutions with Attention for Cross-domain Beauty and Personal Care Product Retrieval

Lin, Zehang; Yang, Zhenguo; Huang, Feitao; Chen, Junhong

doi:10.1145/3240508.3266436

Cited by 28 publications

(26 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Compared with RA-MAC [2], MFF [3], and pretrained ResNet50 [4], which are competitors of Half Million Beauty Product Image Recognition Challenge 2018, our proposed MANet achieves the optimal result on Perfect-500K with 0.395 MAP@7, which is higher than those of RA-MAC with 0.348 MAP@7, MFF with 0.360 MAP@7, and pre-trained ResNet50 with 0.207 MAP@7, respectively.…”

mentioning

confidence: 88%

“…Dear editor, In recent years, the Perfect Half Million Beauty Product Image Recognition Challenge has been held by ACM MultiMedia 2018 [1] for beauty product image retrieval task, and the Perfect-500K dataset has been released, acting as a large-scale beauty product dataset. Retrieval methods exploit classic CNN models to extract features and conduct either fusion or post-process to enhance accuracy of feature description (e.g., [2][3][4]). Others are inclined to design network architecture to achieve the idential effect (e.g., [5][6][7]).…”

mentioning

confidence: 99%

See 1 more Smart Citation

Multi-attention based cross-domain beauty product image retrieval

Wang

Xing

Jia-jun

et al. 2020

Sci. China Inf. Sci.

View full text Add to dashboard Cite

mentioning

confidence: 88%

mentioning

confidence: 99%

Multi-attention based cross-domain beauty product image retrieval

Wang

Xing

Jia-jun

et al. 2020

Sci. China Inf. Sci.

View full text Add to dashboard Cite

“…However, these works concentrate on local descriptors and neglect the global descriptors. To make up the loss, we shift the focus to the global descriptors and determine to modify the MAC descriptor which shows the best performance in beauty retrieval task among global descriptors [5,17]. As shown in Fig.3, given an image ∈ ℎ× ×3 , the backbone firstly takes it as input and maps it to feature maps ∈ ℎ1× 1× ( is the depth of feature maps).…”

Section: Attention-based Maximum Activation Convolutionsmentioning

confidence: 99%

“…Thus, the background will be a disturbance and degrade the performance substantially when applying CBIR descriptors. To tackle this difficulty, some researchers [5,17] introduce attention mechanism into descriptors and obtain favorable performance. However, these modifications focus on local descriptors and neglect the necessity of improving global descriptors with this mechanism.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Attention Based Beauty Product Retrieval Using Global and Local Descriptors

Xie

et al. 2020

Proceedings of the 28th ACM International Conference on Multimedia

View full text Add to dashboard Cite

Beauty product retrieval has drawn more and more attention for its wide application outlook and enormous economic benefits. However, this task is always challenging due to the variation of products, especially the disturbance of clustered background. In this paper, we first introduce attention mechanism into a global image descriptor, i.e., Maximum Activation of Convolutions (MAC), and propose Attention-based MAC (AMAC). With this enhancement, we can suppress the negative effect of background and highlight the foreground in an unsupervised manner. Then, AMAC and local descriptors are ensembled to complementarily increase the performance. Furthermore, we try to finetune multiple retrieval methods on the different datasets and adopt a query expansion strategy to obain more improvements. Extensive experiments conducted on a dataset containing more the half million beauty products (Perfect-500K) demonstrate the effectiveness of the proposed method. Finally, our team (USTC-NELSLIP) wins the first place on the leaderboard of the "AI Meets Beauty" Grand Challenge of ACM Multimedia 2020.

show abstract

Survey on clothing image retrieval with cross-domain

Chen

Yang

Menglu

2022

Complex Intell. Syst.

View full text Add to dashboard Cite

The paper summarizes the research progress on critical region recognition and deep metric learning to achieve accurate clothing image retrieval in cross-domain situations. Critical region recognition is of great value for the clothing feature extraction, effectively improving retrieval accuracy. The accuracy will decrease when solving difficult samples with similar features but different categories. Nowadays, deep metric learning is an effective way to solve this problem, which utilizes the optimization of different loss functions and ensemble network to strengthen the discrimination of clothing features. Therefore, through comparison of the experimental results of different algorithms and analysis of the accuracy of cross-domain clothing retrieval, it is demonstrated that the improvement of the retrieval accuracy in the future mainly depends on clothing important feature extraction and clothing feature discrimination.

show abstract

Regional Maximum Activations of Convolutions with Attention for Cross-domain Beauty and Personal Care Product Retrieval

Cited by 28 publications

References 11 publications

Multi-attention based cross-domain beauty product image retrieval

Multi-attention based cross-domain beauty product image retrieval

Attention Based Beauty Product Retrieval Using Global and Local Descriptors

Survey on clothing image retrieval with cross-domain

Contact Info

Product

Resources

About