Filters








40,131 Hits in 6.9 sec

Metric Learning with Adaptive Density Discrimination [article]

Oren Rippel, Manohar Paluri, Piotr Dollar, Lubomir Bourdev
2016 arXiv   pre-print
Distance metric learning (DML) approaches learn a transformation to a representation space where distance is in correspondence with a predefined notion of similarity.  ...  It then employs this knowledge to adaptively assess similarity, and achieve local discrimination by penalizing class distribution overlap.  ...  Figure 1 : Distance metric learning approaches sculpt a representation space where distance is in correspondence with a notion of similarity.  ... 
arXiv:1511.05939v2 fatcat:xgxd2bn37ne5reckmdjjf3dhr4

Unsupervised Vehicle Counting via Multiple Camera Domain Adaptation [article]

Luca Ciampi and Carlos Santiago and Joao Paulo Costeira and Claudio Gennaro and Giuseppe Amato
2020 arXiv   pre-print
We propose and discuss a new methodology to design image-based vehicle density estimators with few labeled data via multiple camera domain adaptations.  ...  This is a recurrent problem when dealing with physical systems and a key research area in Machine Learning and AI.  ...  Domain Adaptation Learning The proposed framework is trained based on an alternate optimization of density estimation network, Ψ, and the discriminator network, Θ.  ... 
arXiv:2004.09251v2 fatcat:4v73ssk33fd5ri7dl7lufxpobe

Visual discrimination and adaptation using non-linear unsupervised learning

Sandra Jiménez, Valero Laparra, Jesus Malo, Bernice E. Rogowitz, Thrasyvoulos N. Pappas, Huib de Ridder
2013 Human Vision and Electronic Imaging XVIII  
Moreover, these algorithms have to be flexible enough to account for the non-linear and adaptive behavior of the system.  ...  relevant as well, 11 but also, statistical independence may not be the better solution to make optimal inferences in squared error terms. 12-14 Moreover, linear methods cannot account for the non-uniform discrimination  ...  compute the discrimination metric in the stimulus domain as previously done with empirical divisive-normalization models. 21, 22 Moreover, using the local density information SPCA and RBIG can be tuned  ... 
doi:10.1117/12.2019008 dblp:conf/hvei/JimenezLM13 fatcat:ylxkgrqatjgvndh5vot7j7gvpy

Deep Metric Learning with Density Adaptivity [article]

Yehao Li and Ting Yao and Yingwei Pan and Hongyang Chao and Tao Mei
2019 arXiv   pre-print
The problem of distance metric learning is mostly considered from the perspective of learning an embedding space, where the distances between pairs of examples are in correspondence with a similarity metric  ...  With the rise and success of Convolutional Neural Networks (CNN), deep metric learning (DML) involves training a network to learn a nonlinear transformation to the embedding space.  ...  DEEP METRIC LEARNING WITH DENSITY ADAPTIVITY Our proposed Deep Metric Learning with Density Adaptivity (DML-DA) approach is to build an embedding space in which the feature representations of images could  ... 
arXiv:1909.03909v1 fatcat:b4dj5qns3jendd2n2n2cx5nquq

Maximum Density Divergence for Domain Adaptation [article]

Li Jingjing, Chen Erpeng, Ding Zhengming, Zhu Lei, Lu Ke, Shen Heng Tao
2020 IEEE Transactions on Software Engineering   pre-print
In this paper, we propose a new domain adaptation method named Adversarial Tight Match (ATM) which enjoys the benefits of both adversarial training and metric learning.  ...  At last, we tailor the proposed MDD as a practical learning loss and report our ATM.  ...  The state-of-the-art domain adaptation methods align the two domains by focusing on either minimizing a divergence metric or confusing a domain discriminator to learn domaininvariant features.  ... 
doi:10.1109/tpami.2020.2991050 pmid:32356736 arXiv:2004.12615v1 fatcat:necre6lisvb3rkgnzfhr47lvje

Principle of Learning Metrics for Exploratory Data Analysis

Samuel Kaski, Janne Sinkkonen
2004 Journal of VLSI Signal Processing Systems for Signal, Image and Video Technology  
The learning of the metric is supervised by the auxiliary data, whereas the data analysis in the new metric is unsupervised.  ...  Our recently developed methods remove this arbitrariness by learning to measure important differences. The effect is equivalent to changing the metric of the data space.  ...  Self-Organizing Maps in Learning Metrics Learning metrics can be used with several unsupervised and supervised methods.  ... 
doi:10.1023/b:vlsi.0000027483.39774.f8 fatcat:sy22ecjk3bcjnblhz3zkfeqbae

Traffic Density Estimation via Unsupervised Domain Adaptation (Discussion Paper)

Luca Ciampi, Carlos Santiago, João Paulo Costeira, Claudio Gennaro, Giuseppe Amato
2021 Sistemi Evoluti per Basi di Dati  
We propose a new methodology to design image-based vehicle density estimators with few labeled data via an unsupervised domain adaptation technique.  ...  Domain Adaptation Learning The proposed framework is trained based on an alternate optimization of the density estimation network, Ψ, and the discriminator network, Θ.  ...  At the same time, it learns to predict realistic density maps for the target domain by trying to fool the discriminator with an adversarial loss.  ... 
dblp:conf/sebd/CiampiSCGA21 fatcat:denaunsbkfdzxpylxuci37ixbe

Discriminative Clustering: Optimal Contingency Tables by Learning Metrics [chapter]

Janne Sinkkonen, Samuel Kaski, Janne Nikkilä
2002 Lecture Notes in Computer Science  
The learning metrics principle describes a way to derive metrics to the data space from paired data.  ...  In this paper, discriminative clustering using a mutual information criterion is shown to be asymptotically equivalent to vector quantization in learning metrics.  ...  The principle of learning metrics aims at automating part of the process of metric selection, by learning the metric from data.  ... 
doi:10.1007/3-540-36755-1_35 fatcat:q5ag4noterdelmiuzp6yzrgvdu

AD-Cluster: Augmented Discriminative Clustering for Domain Adaptive Person Re-identification [article]

Yunpeng Zhai
2020 arXiv   pre-print
AD-Cluster is trained by iterative density-based clustering, adaptive sample augmentation, and discriminative feature learning.  ...  with the augmented clusters.  ...  With density-based clustering, we introduce adaptive sample augmentation to generate more diverse samples and a min-max optimization scheme to learn more discriminative re-ID model.  ... 
arXiv:2004.08787v2 fatcat:aje7vv3kgzby7iiln7kv2egbaa

CODA: Counting Objects via Scale-aware Adversarial Density Adaption [article]

Li Wang, Yongbo Li, Xiangyang Xue
2019 arXiv   pre-print
adversarial Density Adaption).  ...  Further analysis indicates that our density adaption framework can effortlessly extend to scenarios with different objects. The code is available at https://github.com/Willy0919/CODA.  ...  Density Adaption To perform density adaption process, we reuse the CN as density map generator, and another discriminator is added to carry out adversarial learning during adaption process.  ... 
arXiv:1903.10442v1 fatcat:y7vkltpxvffczkdpjm5ihzt5kq

Coarse to Fine: Domain Adaptive Crowd Counting via Adversarial Scoring Network [article]

Zhikang Zou, Xiaoye Qu, Pan Zhou, Shuangjie Xu, Xiaoqing Ye, Wenhao Wu, Jin Ye
2021 arXiv   pre-print
learning.  ...  In specific, at the coarse-grained stage, we design a dual-discriminator strategy to adapt source domain to be close to the targets from the perspectives of both global and local feature space via adversarial  ...  The is trained using the Stochastic Gradient Descent (SGD) optimizer with a learning rate as 10 −6 . We use Adam optimizer [24] with learning rate of 10 −4 for the discriminators.  ... 
arXiv:2107.12858v1 fatcat:xheysldadrh4hefguvwncnrzuq

AD-Cluster: Augmented Discriminative Clustering for Domain Adaptive Person Re-Identification

Yunpeng Zhai, Shijian Lu, Qixiang Ye, Xuebo Shan, Jie Chen, Rongrong Ji, Yonghong Tian
2020 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
AD-Cluster is trained by iterative densitybased clustering, adaptive sample augmentation, and discriminative feature learning.  ...  with the augmented clusters.  ...  With density-based clustering, we introduce adaptive sample augmentation to generate more diverse samples and a min-max optimization scheme to learn more discriminative re-ID model.  ... 
doi:10.1109/cvpr42600.2020.00904 dblp:conf/cvpr/ZhaiLYSCJ020 fatcat:z7mfvavt6ncjzbb4pgp4s6xfwq

Learning Discriminative Metrics via Generative Models and Kernel Learning [article]

Yuan Shi, Yung-Kyun Noh, Fei Sha, Daniel D. Lee
2011 arXiv   pre-print
In this paper, we show how to unify generative and discriminative learning of metrics via a kernel learning framework.  ...  Metrics specifying distances between data points can be learned in a discriminative manner or from generative models.  ...  Discriminative learning with multiple generative metrics Prior empirical studies have shown that generative learning metric (GLM) of eq. (8) performs competitively, even when compared to discriminative  ... 
arXiv:1109.3940v1 fatcat:totbvwpmmjbyzpanvhacacsptu

Background-Aware Domain Adaptation for Plant Counting

Min Shi, Xing-Yi Li, Hao Lu, Zhi-Guo Cao
2022 Frontiers in Plant Science  
This problem setting is also called unsupervised domain adaptation (UDA). Despite UDA has been a long-standing topic in machine learning society, UDA methods are less studied for plant counting.  ...  We also show that BADA can work with adversarial training strategies to further enhance the robustness of counting models against the domain gap.  ...  According to the proposed metric DMAE, UDA method can also help the model predict more precise density maps.  ... 
doi:10.3389/fpls.2022.731816 pmid:35185973 pmcid:PMC8850787 fatcat:db6gifzyjzhrfdfil57l2pxby4

Rethinking Sampling in 3D Point Cloud Generative Adversarial Networks [article]

He Wang, Zetian Jiang, Li Yi, Kaichun Mo, Hao Su, Leonidas J. Guibas
2020 arXiv   pre-print
Through extensive experiments, we show that sampling-insensitive discriminators (e.g.PointNet-Max) produce shape point clouds with point clustering artifacts while sampling-oversensitive discriminators  ...  metrics.  ...  Different with averaging, the learned correlation coefficients between the points GH T no longer maintains the information of point density distribution, and that's why leveraging a mix-pooling in this  ... 
arXiv:2006.07029v1 fatcat:x3qjxbnqmngfnfoqjkbx4wssdy
« Previous Showing results 1 — 15 out of 40,131 results