Filters








1,139 Hits in 7.3 sec

Intra-class Variation Isolation in Conditional GANs [article]

Richard T. Marriott, Sami Romdhani, Liming Chen
2018 arXiv   pre-print
We coin the method intra-class variation isolation (IVI) and the resulting network the IVI-GAN.  ...  the weak supervision of binary attribute labels.  ...  Intra-class variation isolation Before describing intra-class variation isolation, we introduce the conditional GAN.  ... 
arXiv:1811.11296v1 fatcat:sa6cwtgucjbmzponaryw3vpiyy

Joint Deep Learning of Facial Expression Synthesis and Recognition [article]

Yan Yan, Ying Huang, Si Chen, Chunhua Shen, Hanzi Wang
2020 arXiv   pre-print
to reduce the intra-class variations of images from the same class, which can significantly improve the final performance.  ...  Moreover, in order to alleviate the problem of data bias between the real images and the synthetic images, we propose an intra-class loss with a novel real data-guided back-propagation (RDBP) algorithm  ...  Therefore, except for the intra-class loss defined above, R is also trained under the supervision of the classification loss, which is defined as, L cls,R = L r cls,R + L f cls,R , (19) where L r cls,R  ... 
arXiv:2002.02194v1 fatcat:gtjnfrbk3bg5lidkrengp7f76a

Disentangling factors of variation in deep representations using adversarial training [article]

Michael Mathieu, Junbo Zhao, Pablo Sprechmann, Aditya Ramesh, Yann LeCun
2016 arXiv   pre-print
In both instances, the intra-class diversity is the source of the unspecified factors of variation: each object is observed at multiple viewpoints, and each speaker dictates multiple phrases.  ...  Experimental results on synthetic and real datasets show that the proposed method is capable of generalizing to unseen classes and intra-class variabilities.  ...  Train this system in a plain supervised fashion to learn the class of the samples.  ... 
arXiv:1611.03383v1 fatcat:qhe2gyczu5hpbj2h64s5qusz5y

Pose-Guided Photorealistic Face Rotation

Yibo Hu, Xiang Wu, Bing Yu, Ran He, Zhenan Sun
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
It not only forms a mask image to guide the generator in learning process but also provides a flexible controllable condition during inference.  ...  Besides the generator and conditional adversarial loss, CAPG-GAN further employs identity preserving loss and total variation regularization to preserve identity information and refine local textures respectively  ...  It makes the same subject form a compact cluster with small intra-class distances and variances in embedding space.  ... 
doi:10.1109/cvpr.2018.00876 dblp:conf/cvpr/HuWYHS18 fatcat:trl5znmxznfg7fcyvdvspbauee

Cross-Domain Face Synthesis using a Controllable GAN [article]

Fania Mokhayeri, Kaveh Kamali, Eric Granger
2019 arXiv   pre-print
This allows generating realistic synthetic face images that reflects capture conditions in the target domain while controlling the GAN output to generate faces under desired pose conditions.  ...  Moreover, despite the emergence of Generative Adversarial Networks (GANs) for realistic synthetic generation, it is often difficult to control the conditions under which synthetic faces are generated.  ...  Although their results are encouraging, the synthetic face images may not be realistic enough to represent intra-class variations of target domain capture conditions.  ... 
arXiv:1910.14247v1 fatcat:pvshgzwtn5etfdmgllb32vkplm

Multimodal Co-learning: Challenges, Applications with Datasets, Recent Advances and Future Directions [article]

Anil Rahate, Rahee Walambe, Sheela Ramanna, Ketan Kotecha
2021 arXiv   pre-print
To that end, in this work, we provide a comprehensive survey on the emerging area of multimodal co-learning that has not been explored in its entirety yet.  ...  In the current state of multimodal machine learning, the assumptions are that all modalities are present, aligned, and noiseless during training and testing time.  ...  to influence the work reported in this paper.  ... 
arXiv:2107.13782v2 fatcat:s4spofwxjndb7leqbcqnwbifq4

A Survey on Adversarial Image Synthesis [article]

William Roy, Glen Kelly, Robert Leer, Frederick Ricardo
2021 arXiv   pre-print
Among the many applications of GAN, image synthesis is the most well-studied one, and research in this area has already demonstrated the great potential of using GAN in image synthesis.  ...  as possible future research directions in image synthesis with GAN.  ...  The loss function is optimized to improve the class prediction. Class conditioning is applied in the hidden space to run the generation procedure towards the objected class.  ... 
arXiv:2106.16056v2 fatcat:mivx26q4x5ampfi566tipcwv3e

Cross-Domain Face Synthesis using a Controllable GAN

Fania Mokhayeri, Kaveh Kamali, Eric Granger
2020 2020 IEEE Winter Conference on Applications of Computer Vision (WACV)  
This allows generating realistic synthetic face images that reflect capture conditions in the target domain, while controlling the GAN output such that faces may be generated under desired pose conditions  ...  Moreover, despite the emergence of Generative Adversarial Networks (GANs) for realistic synthetic generation, it is often difficult to control the conditions under which synthetic faces are generated.  ...  Although their results are encouraging, the synthetic face images may not be realistic enough to represent intra-class variations of target domain capture conditions.  ... 
doi:10.1109/wacv45572.2020.9093275 dblp:conf/wacv/MokhayeriKG20 fatcat:7ub5dfmxnzeepjev35fiu73ram

Local Clustering with Mean Teacher for Semi-supervised Learning [article]

Zexi Chen, Benjamin Dutton, Bharathkumar Ramachandra, Tianfu Wu, Ranga Raju Vatsavai
2020 arXiv   pre-print
MT maintains a teacher model's weights as the exponential moving average of a student model's weights and minimizes the divergence between their probability predictions under diverse perturbations of the  ...  We demonstrate on semi-supervised benchmark datasets SVHN and CIFAR-10 that adding our LC loss to MT yields significant improvements compared to MT and performance comparable to the state of the art in  ...  Instead of employing class labels for supervision, they define weak labels in the form of similar/dissimilar data pairs, so that the clustering strategy is to minimize pairwise distances between similar  ... 
arXiv:2004.09665v2 fatcat:uk766o5ynzhmdhudtk3gziyiie

ContraGAN: Contrastive Learning for Conditional Image Generation [article]

Minguk Kang, Jaesik Park
2021 arXiv   pre-print
Conditional image generation is the task of generating diverse images using class label information.  ...  In this paper, we propose ContraGAN that considers relations between multiple image embeddings in the same batch (data-to-data relations) as well as the data-to-class relations by using a conditional contrastive  ...  On the other hand, 2C loss utilizes weak supervision of label information.  ... 
arXiv:2006.12681v3 fatcat:tlnifrdon5carcvihhtbytedpa

Probabilistic Video Generation using Holistic Attribute Control [article]

Jiawei He, Andreas Lehrmann, Joseph Marino, Greg Mori, Leonid Sigal
2018 arXiv   pre-print
Variational Autoencoders (VAEs) are used as a means of encoding/decoding frames into/from the latent space and RNN as a wayto model the dynamics in the latent space.  ...  We improve the video generation consistency through temporally-conditional sampling and quality by structuring the latent space with attribute controls; ensuring that attributes can be both inferred and  ...  attribute control (Sec. 3.2) and a conditional variational posterior (Sec. 3.3).  ... 
arXiv:1803.08085v1 fatcat:nzylgu2tyjgr7frm5ismzoze5i

Deep Visual Domain Adaptation: A Survey [article]

Mei Wang, Weihong Deng
2018 arXiv   pre-print
In this paper, we provide a comprehensive survey of deep domain adaptation methods for computer vision applications with four major contributions.  ...  Second, we summarize deep domain adaption approaches into several categories based on training loss, and analyze and compare briefly the state-of-the-art methods under these categories.  ...  In contrast to other works in which the generator is conditioned only on a noise vector or source images, Bousmalis et al. [4] proposed a model that exploits GANs conditioned on both.  ... 
arXiv:1802.03601v4 fatcat:d5hwwecipjfjzmh7725lmepzfe

An Overview of Deep Semi-Supervised Learning [article]

Yassine Ouali, Céline Hudelot, Myriam Tami
2020 arXiv   pre-print
In this paper, we provide a comprehensive overview of deep semi-supervised learning, starting with an introduction to the field, followed by a summarization of the dominant semi-supervised approaches in  ...  In a search for more data-efficient deep learning methods to overcome the need for large annotated datasets, there is a rising research interest in semi-supervised learning and its applications to deep  ...  Ouali is supported by Randstad corporate research in collaboration with Université Paris-Saclay, Cen-traleSupélec, MICS. We thank Victor Bouvier for his helpful feedback on an earlier version.  ... 
arXiv:2006.05278v2 fatcat:gvxqpel3xnhejlq3yqktynqvim

A Survey on Face Data Augmentation [article]

Xiang Wang and Kai Wang and Shiguo Lian
2019 arXiv   pre-print
In this paper, we systematically review the existing works of face data augmentation from the perspectives of the transformation types and methods, with the state-of-the-art approaches involved.  ...  We point out the challenges and opportunities in the field of face data augmentation, and provide brief yet insightful discussions.  ...  In comparison with IS, FID is more robust to noise and more sensitive to intra-class mode collapse [139] .  ... 
arXiv:1904.11685v1 fatcat:phcwwc7gcfablgytt6itr6xade

A multi-stage semi-supervised improved deep embedded clustering method for bearing fault diagnosis under the situation of insufficient labeled samples [article]

Tongda Sun, Gang Yu
2021 arXiv   pre-print
This method provides a new approach for fault diagnosis under the situation of limited labeled samples by effectively exploring unsupervised data.  ...  However, a difficulty of implementing this in real industries hinders the application of these methods.  ...  In this section, experimental comparisons of different methods prove the effectiveness and advancement of the proposed method in the task of bearing fault diagnosis under weak supervision.  ... 
arXiv:2109.13521v2 fatcat:co2qgdrqezeirnpts5un4zn2z4
« Previous Showing results 1 — 15 out of 1,139 results