GeoStyle: Discovering Fashion Trends and Events---GeoStyle:发现时尚趋势和事件

-	-
GeoStyle: Discovering Fashion Trends and Events---ICCV 2019 Open Access Repository

GeoStyle: Discovering Fashion Trends and Events
GeoStyle:发现时尚趋势和事件

Abstract
摘要

Understanding fashion styles and trends is of great potential interest to retailers and consumers alike.
了解时尚风格和趋势对零售商和消费者都有很大的潜在利益。
The photos people upload to social media are a historical and public data source of how people dress across the world and at different times.
人们上传到社交媒体上的照片是一个历史和公共数据来源，反映了世界各地不同时期人们的着装。
While we now have tools to automatically recognize the clothing and style attributes of what people are wearing in these photographs, we lack the ability to analyze spatial and temporal trends in these attributes or make predictions about the future.
虽然我们现在有了自动识别人们在这些照片中穿着的服装和风格属性的工具，但我们缺乏分析这些属性的时空趋势或预测未来的能力。
In this paper, we address this need by providing an automatic framework that analyzes large corpora of street imagery to (a) discover and forecast long-term trends of various fashion attributes as well as automatically discovered styles, and (b) identify spatio-temporally localized events that affect what people wear.
在本文中，我们通过提供一个自动的框架来解决这一需求，该框架可以分析大量的街道图像，从而(a)发现和预测各种时尚属性的长期趋势以及自动发现的风格，(b)识别影响人们穿着的时空本地化事件。
We show that our framework makes long term trend forecasts that are >20 identifies hundreds of socially meaningful events that impact fashion across the globe.
我们展示了我们的框架做出的>20的长期趋势预测，确定了数百个影响全球时尚的有社会意义的事件。

1INTRODUCTION
1介绍
Each day, we collectively upload to social media platforms billions of photographs that capture a wide range of human life and activities around the world.
每天，我们都会向社交媒体平台上传数十亿张照片，这些照片记录了世界各地人们的生活和活动。
At the same time, object detection, semantic segmentation, and visual search are seeing rapid advances [13] and are being deployed at scale [22].
与此同时，对象检测、语义分割和视觉搜索也得到了快速的发展，并在[22]的规模上得到了部署。
With large-scale recognition available as a fundamental tool in our vision toolbox, it is now possible to ask questions about how people dress, eat, and group across the world and over time.
随着大规模识别成为我们视觉工具箱中的基本工具，现在我们可以询问世界各地人们的穿着、饮食和群体状况。
In this paper we focus on how people dress.
在这篇论文中，我们关注的是人们的穿着。
In particular, we ask can we detect and predict fashion trends and styles over space and timeWe answer these questions by designing an automated method to characterize and predict seasonal and year-over-year fashion trends, detect social events (e.g., festivals or sporting events) that impact how people dress, and identify social-event-specific style elements that epitomize these events.
特别是,我们问我们能检测和预测流行趋势和风格在空间和timeWe回答这些问题设计了一个自动化的方法来描述和预测季节性和时尚趋势,同比检测社交活动(例如,节日或体育赛事),影响人们如何着装,并确定social-event-specific风格元素概括这些事件。
Our approach uses existing recognition algorithms to identify a coarse set of fashion attributes in a large corpus of images.
我们的方法使用现有的识别算法来识别大量图像中的一组粗糙的时尚属性。
We then fit interpretable parametric models of long-term temporal trends to these fashion attributes.
然后，我们将可解释的长期时间趋势参数模型与这些时尚属性进行拟合。
These models capture both seasonal cycles as well as changes in popularity over time.
这些模型捕捉了季节周期和流行度随时间的变化。
These models not only help in understanding existing trends, but can also make up to 20% more accurate, temporally fine-grained forecasts across long time scales compared to prior methods [1].
这些模型不仅有助于理解现有的趋势，而且与以前的[1]方法相比，它们还可以在较长时间范围内提供高达20%的准确性、时间上的细粒度预测。
For example, we find that year-on-year more people are wearing black, but that they tend to do so more in the winter than in the summer.
例如，我们发现，与去年同期相比，穿黑色衣服的人越来越多，但冬季穿黑色衣服的人往往比夏季多。
Our framework not only models long-term trends, but also identifies sudden, short-term changes in popularity that buck these trends.
我们的框架不仅模拟长期趋势，而且还识别出与这些趋势相反的突然的、短期的流行变化。
We find that these outliers often correspond to festivals, sporting events, or other large social gatherings.
我们发现这些异常值通常对应于节日、体育赛事或其他大型社交聚会。
We provide a methodology to automaticallydiscover the events underlying such outliers by looking at associated image tags and captions, thus tying visual analysis to text-based discovery.
我们提供了一种方法，通过查看相关的图像标记和标题，自动发现这些异常值背后的事件，从而将可视化分析与基于文本的发现结合起来。
We find that our framework finds understandable reasons for all of the most salient events it discovers, and in so doing surfaces intriguing social events around the world that were unknown to the authors.
我们发现，我们的框架为它所发现的所有最显著的事件找到了可以理解的原因，并在这样做的过程中，呈现出作者所不知道的世界各地有趣的社会事件。
For example, it discovers an unusual increase in the color yellow in Bangkok in early December, and associates it with the words father, day, king, live, and dad.
例如，它在12月初发现曼谷的黄色有所增加，并将其与“父亲”、“日子”、“国王”、“生活”和“爸爸”联系起来。
This corresponds to the kings birthday, celebrated as Fathers Day in Thailand by wearing yellow [36].
这一天正好是国王的生日，在泰国，人们穿着黄色的[36]来庆祝父亲节。
Our framework similarly surfaces events in Ukraine (Vyshyvanka Day), Indonesia (Batik Day), and Japan (Golden Week).
我们的框架同样处理乌克兰(Vyshyvanka日)、印度尼西亚(蜡染日)和日本(黄金周)的事件。
Figure 1 shows more of the worldwide events discovered by our framework and the clothes that people wear during those events.
图1显示了我们的框架发现的更多的全球事件，以及人们在这些事件中穿的衣服。
Figure 1 Major events discovered by our framework.
图1我们的框架发现的主要事件。
For each event, the figure shows the clothing that people typically wear for that event, along with the city, one of the months of occurrence, and the most descriptive word extracted using the images captions.
对于每个事件，图中显示了人们通常在该事件中穿的衣服，以及事件发生的一个月所在的城市，以及使用图像说明提取的最具描述性的单词。
The inset image shows more precise locations of these cities.
插图显示了这些城市更精确的位置。
We further show that we can predict trends and events not just at the level of individual fashion attributes (such as wearing yellow), but also at the level of styles consisting of recurring visual ensembles.
我们进一步证明，我们不仅可以在个人时尚属性(如穿黄色衣服)的层次上预测趋势和事件，而且还可以在由反复出现的视觉效果组成的风格层次上预测趋势和事件。
These styles are identified by clustering photographs in feature space to reveal style clusters clusters of people dressed in a similar style.
这些风格是通过在特征空间中聚集照片来识别的，以揭示风格集群，集群的人穿着相似的风格。
Our forecasts of the future popularity of styles are just as accurate as our predictions of individual attributes.
我们对未来流行风格的预测和我们对个人属性的预测一样准确。
Further, we can run the same event detection framework described above on style trends, allowing us to not only automatically detect social events, but also associate each event with its own distinctive style a stylistic signature for each event.
此外，我们还可以在style trends上运行与上述相同的事件检测框架，不仅可以自动检测社会事件，还可以将每个事件与它自己的独特风格相关联，即每个事件的风格特征。
Our contributions, highlighted in Figure 2, includeWe present an automated framework for analyzing the temporal behavior of fashion elements across the globe.
我们的贡献，在图2中突出显示，包括我们提供了一个自动化的框架来分析全球各地的时尚元素的临时行为。
Our framework models and forecasts long-term trends and seasonal behaviors.
我们的框架模型和预测长期趋势和季节行为。
It also automatically identifies short-term spikes caused by events like festivals and sporting events.
它还自动识别由节日和体育赛事等事件造成的短期峰值。
Our framework automatically discovers the reasons behind these events by leveraging textual descriptions and captions.
我们的框架通过利用文本描述和说明自动发现这些事件背后的原因。
We connect events with signature styles by performing this analysis on automatically discovered style clusters.
通过对自动发现的样式集群执行此分析，我们将事件与签名样式连接起来。

---

GeoStyle: Discovering Fashion Trends and Events

Abstract

Understanding fashion styles and trends is of great potential interest to retailers and consumers alike. The photos people upload to social media are a historical and public data source of how people dress across the world and at different times. While we now have tools to automatically recognize the clothing and style attributes of what people are wearing in these photographs, we lack the ability to analyze spatial and temporal trends in these attributes or make predictions about the future. In this paper, we address this need by providing an automatic framework that analyzes large corpora of street imagery to (a) discover and forecast long-term trends of various fashion attributes as well as automatically discovered styles, and (b) identify spatio-temporally localized events that affect what people wear. We show that our framework makes long term trend forecasts that are >20 identifies hundreds of socially meaningful events that impact fashion across the globe.

1INTRODUCTION

Each day, we collectively upload to social media platforms billions of photographs that capture a wide range of human life and activities around the world. At the same time, object detection, semantic segmentation, and visual search are seeing rapid advances [13] and are being deployed at scale [22]. With large-scale recognition available as a fundamental tool in our vision toolbox, it is now possible to ask questions about how people dress, eat, and group across the world and over time. In this paper we focus on how people dress. In particular, we ask can we detect and predict fashion trends and styles over space and timeWe answer these questions by designing an automated method to characterize and predict seasonal and year-over-year fashion trends, detect social events (e.g., festivals or sporting events) that impact how people dress, and identify social-event-specific style elements that epitomize these events. Our approach uses existing recognition algorithms to identify a coarse set of fashion attributes in a large corpus of images. We then fit interpretable parametric models of long-term temporal trends to these fashion attributes. These models capture both seasonal cycles as well as changes in popularity over time. These models not only help in understanding existing trends, but can also make up to 20% more accurate, temporally fine-grained forecasts across long time scales compared to prior methods [1]. For example, we find that year-on-year more people are wearing black, but that they tend to do so more in the winter than in the summer.Our framework not only models long-term trends, but also identifies sudden, short-term changes in popularity that buck these trends. We find that these outliers often correspond to festivals, sporting events, or other large social gatherings. We provide a methodology to automaticallydiscover the events underlying such outliers by looking at associated image tags and captions, thus tying visual analysis to text-based discovery. We find that our framework finds understandable reasons for all of the most salient events it discovers, and in so doing surfaces intriguing social events around the world that were unknown to the authors. For example, it discovers an unusual increase in the color yellow in Bangkok in early December, and associates it with the words father, day, king, live, and dad. This corresponds to the kings birthday, celebrated as Fathers Day in Thailand by wearing yellow [36]. Our framework similarly surfaces events in Ukraine (Vyshyvanka Day), Indonesia (Batik Day), and Japan (Golden Week). Figure 1 shows more of the worldwide events discovered by our framework and the clothes that people wear during those events.Figure 1 Major events discovered by our framework. For each event, the figure shows the clothing that people typically wear for that event, along with the city, one of the months of occurrence, and the most descriptive word extracted using the images captions. The inset image shows more precise locations of these cities.We further show that we can predict trends and events not just at the level of individual fashion attributes (such as wearing yellow), but also at the level of styles consisting of recurring visual ensembles. These styles are identified by clustering photographs in feature space to reveal style clusters clusters of people dressed in a similar style. Our forecasts of the future popularity of styles are just as accurate as our predictions of individual attributes. Further, we can run the same event detection framework described above on style trends, allowing us to not only automatically detect social events, but also associate each event with its own distinctive style a stylistic signature for each event.Our contributions, highlighted in Figure 2, includeWe present an automated framework for analyzing the temporal behavior of fashion elements across the globe. Our framework models and forecasts long-term trends and seasonal behaviors. It also automatically identifies short-term spikes caused by events like festivals and sporting events.Our framework automatically discovers the reasons behind these events by leveraging textual descriptions and captions.We connect events with signature styles by performing this analysis on automatically discovered style clusters.

---

GeoStyle:发现时尚趋势和事件

摘要

了解时尚风格和趋势对零售商和消费者都有很大的潜在利益。人们上传到社交媒体上的照片是一个历史和公共数据来源，反映了世界各地不同时期人们的着装。虽然我们现在有了自动识别人们在这些照片中穿着的服装和风格属性的工具，但我们缺乏分析这些属性的时空趋势或预测未来的能力。在本文中，我们通过提供一个自动的框架来解决这一需求，该框架可以分析大量的街道图像，从而(a)发现和预测各种时尚属性的长期趋势以及自动发现的风格，(b)识别影响人们穿着的时空本地化事件。我们展示了我们的框架做出的>20的长期趋势预测，确定了数百个影响全球时尚的有社会意义的事件。

1介绍
每天，我们都会向社交媒体平台上传数十亿张照片，这些照片记录了世界各地人们的生活和活动。与此同时，对象检测、语义分割和视觉搜索也得到了快速的发展，并在[22]的规模上得到了部署。随着大规模识别成为我们视觉工具箱中的基本工具，现在我们可以询问世界各地人们的穿着、饮食和群体状况。在这篇论文中，我们关注的是人们的穿着。特别是,我们问我们能检测和预测流行趋势和风格在空间和timeWe回答这些问题设计了一个自动化的方法来描述和预测季节性和时尚趋势,同比检测社交活动(例如,节日或体育赛事),影响人们如何着装,并确定social-event-specific风格元素概括这些事件。我们的方法使用现有的识别算法来识别大量图像中的一组粗糙的时尚属性。然后，我们将可解释的长期时间趋势参数模型与这些时尚属性进行拟合。这些模型捕捉了季节周期和流行度随时间的变化。这些模型不仅有助于理解现有的趋势，而且与以前的[1]方法相比，它们还可以在较长时间范围内提供高达20%的准确性、时间上的细粒度预测。例如，我们发现，与去年同期相比，穿黑色衣服的人越来越多，但冬季穿黑色衣服的人往往比夏季多。我们的框架不仅模拟长期趋势，而且还识别出与这些趋势相反的突然的、短期的流行变化。我们发现这些异常值通常对应于节日、体育赛事或其他大型社交聚会。我们提供了一种方法，通过查看相关的图像标记和标题，自动发现这些异常值背后的事件，从而将可视化分析与基于文本的发现结合起来。我们发现，我们的框架为它所发现的所有最显著的事件找到了可以理解的原因，并在这样做的过程中，呈现出作者所不知道的世界各地有趣的社会事件。例如，它在12月初发现曼谷的黄色有所增加，并将其与“父亲”、“日子”、“国王”、“生活”和“爸爸”联系起来。这一天正好是国王的生日，在泰国，人们穿着黄色的[36]来庆祝父亲节。我们的框架同样处理乌克兰(Vyshyvanka日)、印度尼西亚(蜡染日)和日本(黄金周)的事件。图1显示了我们的框架发现的更多的全球事件，以及人们在这些事件中穿的衣服。图1我们的框架发现的主要事件。对于每个事件，图中显示了人们通常在该事件中穿的衣服，以及事件发生的一个月所在的城市，以及使用图像说明提取的最具描述性的单词。插图显示了这些城市更精确的位置。我们进一步证明，我们不仅可以在个人时尚属性(如穿黄色衣服)的层次上预测趋势和事件，而且还可以在由反复出现的视觉效果组成的风格层次上预测趋势和事件。这些风格是通过在特征空间中聚集照片来识别的，以揭示风格集群，集群的人穿着相似的风格。我们对未来流行风格的预测和我们对个人属性的预测一样准确。此外，我们还可以在style trends上运行与上述相同的事件检测框架，不仅可以自动检测社会事件，还可以将每个事件与它自己的独特风格相关联，即每个事件的风格特征。我们的贡献，在图2中突出显示，包括我们提供了一个自动化的框架来分析全球各地的时尚元素的临时行为。我们的框架模型和预测长期趋势和季节行为。它还自动识别由节日和体育赛事等事件造成的短期峰值。我们的框架通过利用文本描述和说明自动发现这些事件背后的原因。通过对自动发现的样式集群执行此分析，我们将事件与签名样式连接起来。