没有合适的资源?快使用搜索试试~ 我知道了~
Cross-Domain Similarity Learning for Face Recognition .pdf
1.该资源内容由用户上传,如若侵权请联系客服进行举报
2.虚拟产品一经售出概不退款(资源遇到问题,请及时私信上传者)
2.虚拟产品一经售出概不退款(资源遇到问题,请及时私信上传者)
版权申诉
0 下载量 93 浏览量
2021-11-16
14:37:35
上传
评论
收藏 802KB PDF 举报
温馨提示
![preview](https://dl-preview.csdnimg.cn/43008210/0001-fbb89f9bb22ab34e429d301bfcdfa252_thumbnail.jpeg)
![preview-icon](https://csdnimg.cn/release/downloadcmsfe/public/img/scale.ab9e0183.png)
试读
10页
Cross-Domain Similarity Learning for Face Recognition in Unseen Domains 反事实零样本和开放集视觉识别
资源推荐
资源详情
资源评论
![pdf](https://img-home.csdnimg.cn/images/20210720083512.png)
![pdf](https://img-home.csdnimg.cn/images/20210720083512.png)
![pdf](https://img-home.csdnimg.cn/images/20210720083512.png)
![pdf](https://img-home.csdnimg.cn/images/20210720083512.png)
![rar](https://img-home.csdnimg.cn/images/20210720083606.png)
![pdf](https://img-home.csdnimg.cn/images/20210720083512.png)
![jar](https://img-home.csdnimg.cn/images/20210720083455.png)
![whl](https://img-home.csdnimg.cn/images/20210720083646.png)
![pdf](https://img-home.csdnimg.cn/images/20210720083512.png)
![whl](https://img-home.csdnimg.cn/images/20210720083646.png)
![zip](https://img-home.csdnimg.cn/images/20210720083736.png)
![pdf](https://img-home.csdnimg.cn/images/20210720083512.png)
![zip](https://img-home.csdnimg.cn/images/20210720083736.png)
![zip](https://img-home.csdnimg.cn/images/20210720083736.png)
![exe](https://img-home.csdnimg.cn/images/20210720083343.png)
![](https://csdnimg.cn/release/download_crawler_static/43008210/bg1.jpg)
Cross-Domain Similarity Learning for Face Recognition in Unseen Domains
Masoud Faraki
1
Xiang Yu
1
Yi-Hsuan Tsai
1
Yumin Suh
1
Manmohan Chandraker
1,2
1
NEC Labs America
2
University of California, San Diego
{mfaraki, xiangyu, ytsai, yumin, manu}@nec-labs.com
Abstract
Face recognition models trained under the assumption of
identical training and test distributions often suffer from poor
generalization when faced with unknown variations, such
as a novel ethnicity or unpredictable individual make-ups
during test time. In this paper, we introduce a novel cross-
domain metric learning loss, which we dub Cross-Domain
Triplet (CDT) loss, to improve face recognition in unseen
domains. The CDT loss encourages learning semantically
meaningful features by enforcing compact feature clusters
of identities from one domain, where the compactness is
measured by underlying similarity metrics that belong to
another training domain with different statistics. Intuitively,
it discriminatively correlates explicit metrics derived from
one domain, with triplet samples from another domain in a
unified loss function to be minimized within a network, which
leads to better alignment of the training domains. The net-
work parameters are further enforced to learn generalized
features under domain shift, in a model-agnostic learning
pipeline. Unlike the recent work of Meta Face Recogni-
tion [
18
], our method does not require careful hard-pair
sample mining and filtering strategy during training. Exten-
sive experiments on various face recognition benchmarks
show the superiority of our method in handling variations,
compared to baseline and the state-of-the-art methods.
1. Introduction
Face recognition using deep neural networks has shown
promising outcomes on popular evaluation benchmarks
[
21
,
25
,
26
,
23
]. Many current methods base their ap-
proaches on the assumption that the training data – CASIA-
Webface [
46
] or MS-Celeb-1M [
19
] being the widely used
ones – and the testing data have similar distributions. How-
ever, when deployed to real-world scenarios, those models
often do not generalize well to test data with unknown statis-
tics. In face recognition applications, this may mean a shift
in attributes such as ethnicity, gender or age between the
training and evaluation data. On the other hand, collecting
and labelling more data along the underrepresented attributes
Figure 1. Comparison between the conventional triplet and our
Cross-Domain Triplet losses.
Top:
The standard triplet loss is
domain agnostic and utilizes a shared metric matrix,
Σ
, to measure
distances of all (anchor,positive) and (anchor,negative) pairs.
Bot-
tom:
Our proposed Cross-Domain Triplet loss, takes into account
Σ
+
and
Σ
−
, i.e., the similarity metrics obtained from positive and
negative pairs in one domain, to make compact clusters of triplets
that belong to another domain. This, results to better alignment of
the two domains. Here, colors indicate domains.
is costly. Therefore, given existing data, learning algorithms
are needed to yield universal face representations and in turn,
be applicable across such diverse scenarios.
Domain generalization has recently emerged to address
the same challenge, but mainly for object classification with
limited number of classes [
3
,
9
,
32
]. It aims to employ
multiple labeled source domains with different distributions
to learn a model that generalizes well to unseen target data
at test time. However, many domain generalization methods
are tailored to closed-set scenarios and hence, not directly
applicable if the label spaces of the domains are disjoint.
Generalized face recognition is indeed a prominent example
of open-set applications with very large number of categories,
encouraging the need for further research in this area.
In this paper, we introduce an approach to improve the
problem of face recognition from unseen domains by learn-
ing semantically meaningful representations. To this end,
15292
资源评论
![avatar-default](https://csdnimg.cn/release/downloadcmsfe/public/img/lazyLogo2.1882d7f4.png)
![avatar](https://profile-avatar.csdnimg.cn/ace77722cc904668be9c7ee0feb247ba_dwf1354046363.jpg!1)
![avatar-vip](https://csdnimg.cn/release/downloadcmsfe/public/img/user-vip.1c89f3c5.png)
易小侠
- 粉丝: 6509
- 资源: 9万+
上传资源 快速赚钱
我的内容管理 展开
我的资源 快来上传第一个资源
我的收益
登录查看自己的收益我的积分 登录查看自己的积分
我的C币 登录后查看C币余额
我的收藏
我的下载
下载帮助
![voice](https://csdnimg.cn/release/downloadcmsfe/public/img/voice.245cc511.png)
![center-task](https://csdnimg.cn/release/downloadcmsfe/public/img/center-task.c2eda91a.png)
安全验证
文档复制为VIP权益,开通VIP直接复制
![dialog-icon](https://csdnimg.cn/release/downloadcmsfe/public/img/green-success.6a4acb44.png)