ViBE: Dressing for Diverse Body Shapes

Wei-Lin Hsiao (蕭瑋琳)

[0]

Kristen Grauman

[0]

CVPR, pp. 11056-11066, 2020.

被引用：0|引用|浏览53|DOI:https://doi.org/10.1109/CVPR42600.2020.01107

其它链接：dblp.uni-trier.de|academic.microsoft.com

关键词：

body awaresize recommendationbody shapedifferent bodycollaborative filtering更多(3+)

微博一下：

We explored clothing recommendations that complement an individual’s body shape

摘要：

Body shape plays an important role in determining what garments will best suit a given person, yet today\u0027s clothing recommendation methods take a \"one shape fits all\" approach. These body-agnostic vision methods and datasets are a barrier to inclusion, ill-equipped to provide good suggestions for diverse body shapes. We introduce V...更多

代码：

数据：

精读

下载 PDF 全文

PPT (上传PPT)

简介

Exciting recent advances can link street photos to catalogs [47,54], recommend garments to complete a look [25,33,34,40,73,76], discover styles and trends [3, 32, 57], and search based on subtle visual properties [22, 46]
All such directions promise to augment and accelerate the clothing shopping experience, providing consumers with personalized recommendations and putting a content-based index of products at their fingertips.
Current largescale datasets are heavily biased to a narrow set of body shapes1—typically thin and tall, owing to the fashionista or celebrity photos from which they are drawn [26, 51, 55, 66, 1not to mention skin tone, age, gender, and other demographic factors

重点内容

Research in computer vision is poised to transform the world of consumer fashion
Current largescale datasets are heavily biased to a narrow set of body shapes1—typically thin and tall, owing to the fashionista or celebrity photos from which they are drawn [26, 51, 55, 66, 1not to mention skin tone, age, gender, and other demographic factors
We propose ViBE, a VIsual Body-aware Embedding that captures clothing’s affinity with different body shapes
We explore a new rich online catalog dataset comprised of models of diverse body shape
We explored clothing recommendations that complement an individual’s body shape
We identified a novel source of Web photo data containing fashion models of diverse body shapes, and developed a body-aware embedding to capture clothing’s affinity with different bodies

方法

All methods perform better on dresses than tops
This may be due to the fact that dresses cover a larger portion of the body, and could be inherently more selective about which bodies are suitable.
The more selective or body-specific a garment is, the more value a body-aware recommendation system can offer; the more body-versatile a garment is, the less impact an intelligent recommendation can have.
As the authors focus on the body-specific garments the body-aware embedding’s gain

结论

The authors explored clothing recommendations that complement an individual’s body shape.
The authors identified a novel source of Web photo data containing fashion models of diverse body shapes, and developed a body-aware embedding to capture clothing’s affinity with different bodies.
Through quantitative measurements and human judgments, the authors verified the model’s effectiveness over body-agnostic models, the status quo in the literature.
The authors plan to incorporate the body-aware embedding to address fashion styling and compatibility tasks

总结

Introduction:
Exciting recent advances can link street photos to catalogs [47,54], recommend garments to complete a look [25,33,34,40,73,76], discover styles and trends [3, 32, 57], and search based on subtle visual properties [22, 46]
All such directions promise to augment and accelerate the clothing shopping experience, providing consumers with personalized recommendations and putting a content-based index of products at their fingertips.
Current largescale datasets are heavily biased to a narrow set of body shapes1—typically thin and tall, owing to the fashionista or celebrity photos from which they are drawn [26, 51, 55, 66, 1not to mention skin tone, age, gender, and other demographic factors
Methods:
All methods perform better on dresses than tops
This may be due to the fact that dresses cover a larger portion of the body, and could be inherently more selective about which bodies are suitable.
The more selective or body-specific a garment is, the more value a body-aware recommendation system can offer; the more body-versatile a garment is, the less impact an intelligent recommendation can have.
As the authors focus on the body-specific garments the body-aware embedding’s gain
Conclusion:
The authors explored clothing recommendations that complement an individual’s body shape.
The authors identified a novel source of Web photo data containing fashion models of diverse body shapes, and developed a body-aware embedding to capture clothing’s affinity with different bodies.
Through quantitative measurements and human judgments, the authors verified the model’s effectiveness over body-agnostic models, the status quo in the literature.
The authors plan to incorporate the body-aware embedding to address fashion styling and compatibility tasks

表格

Table1: Dataset statistics: number of garments and fashion models for each clustered type
Table2: Recommendation AUC on unseen people paired with garments sampled from the entire dataset, where ground-truth labels are provided by human judges. Consistent with Fig. 7a, the proposed model outperforms all the baselines

Download tables as Excel

ViBE: Dressing for Diverse Body Shapes

Introduction:

Methods:

Conclusion: