1,715 Hits in 6.0 sec

Overcoming challenges in leveraging GANs for few-shot data augmentation [article]

Christopher Beckham, Issam Laradji, Pau Rodriguez, David Vazquez, Derek Nowrouzezahrai, Christopher Pal
2022 arXiv   pre-print
In this paper, we explore the use of GAN-based few-shot data augmentation as a method to improve few-shot classification performance.  ...  We perform an exploration into how a GAN can be fine-tuned for such a task (one of which is in a class-incremental manner), as well as a rigorous empirical investigation into how well these models can  ...  In this paper we specifically focus on the data aspect of few-shot learning, which at its core leverages some form of data augmentation to augment a small set of examples into a much larger set, one that  ... 
arXiv:2203.16662v2 fatcat:mjekh5q6nreyhkdeuclrhjnbx4

Data-Efficient GAN Training Beyond (Just) Augmentations: A Lottery Ticket Perspective [article]

Tianlong Chen, Yu Cheng, Zhe Gan, Jingjing Liu, Zhangyang Wang
2021 arXiv   pre-print
, ImageNet, and multiple few-shot generation datasets).  ...  Training generative adversarial networks (GANs) with limited real image data generally results in deteriorated performance and collapsed models.  ...  Conclusion and Discussion of Broader Impact We introduce a novel perspective for data-efficient GAN training by leveraging lottery tickets, which augmentations can further enhance, including our newly  ... 
arXiv:2103.00397v3 fatcat:nx67rbomnnhbxeezreqidba6s4

Conservative Generator, Progressive Discriminator: Coordination of Adversaries in Few-shot Incremental Image Synthesis [article]

Chaerin Kong, Nojun Kwak
2022 arXiv   pre-print
To effectively handle the inherent challenges of incremental learning and few-shot learning, we propose a novel framework named ConPro that leverages the two-player nature of GANs.  ...  Several works have previously explored incremental few-shot learning, a task with greater challenges due to data constraint, mostly in classification setting with mild success.  ...  As it differs from simple IL and few-shot learning, we propose a novel GAN framework named ConPro that collaborately addresses the challenges.  ... 
arXiv:2207.14491v1 fatcat:k22bc5hsvvaajhoooltjqwgatu

Few-shot Image Generation with Mixup-based Distance Learning [article]

Chaerin Kong, Jeesoo Kim, Donghoon Han, Nojun Kwak
2022 arXiv   pre-print
In this work, we consider a challenging task of pretraining-free few-shot image synthesis, and seek to train existing generative models with minimal overfitting and mode collapse.  ...  GANs trained with limited data can easily memorize few training samples and display undesirable properties like "stairlike" latent space where interpolation in the latent space yields discontinuous transitions  ...  As naive training of GANs with small datasets often fails both in terms of fidelity and diversity, many have proposed novel approaches specifically designed for few-shot image synthesis.  ... 
arXiv:2111.11672v2 fatcat:eksda5na4jatdlyu47qyjqzqti

Generalized Few-Shot Video Classification with Video Retrieval and Feature Generation [article]

Yongqin Xian, Bruno Korbar, Matthijs Douze, Lorenzo Torresani, Bernt Schiele, Zeynep Akata
2021 arXiv   pre-print
Few-shot learning aims to recognize novel classes from a few examples. Although significant progress has been made in the image domain, few-shot video classification is relatively unexplored.  ...  Moreover, we find existing benchmarks are limited because they only focus on 5 novel classes in each testing episode and introduce more realistic benchmarks by involving more novel classes, i.e. few-shot  ...  In the few-shot learning stage, the same data augmentation is applied and the novel class classifier is learned with a constant learning rate 0.01 for 10 epochs on all the datasets.  ... 
arXiv:2007.04755v2 fatcat:gt4lhkxwefa2bfppqdpdppysle

Few-shot Classifier GAN

Adamu Ali-Gombe, Eyad Elyan, Yann Savoye, Chrisina Jayne
2018 2018 International Joint Conference on Neural Networks (IJCNN)  
32] ✗ CC-GAN [10] ✗ SS-GAN [33] ✗ TAC-GAN [9] ✗ ✗ Few-shot C-GAN (Our) GAN has a lot of applications targeted for images processing, such as image data augmentation [25] , highresolution image  ...  In this paper, we present Few-shot Classifier Generative Adversarial Network as an approach for few-shot classification.  ... 
doi:10.1109/ijcnn.2018.8489387 dblp:conf/ijcnn/Ali-GombeESJ18 fatcat:a3jdkolw2repzmqmhmt55c4yue

Few-Shot Learning with Intra-Class Knowledge Transfer [article]

Vivek Roy, Yan Xu, Yu-Xiong Wang, Kris Kitani, Ruslan Salakhutdinov, Martial Hebert
2020 arXiv   pre-print
Recent works have proposed to solve this task by augmenting the training data of the few-shot classes using generative models with the few-shot training samples as the seeds.  ...  However, due to the limited number of the few-shot seeds, the generated samples usually have small diversity, making it difficult to train a discriminative classifier for the few-shot classes.  ...  Second, we leverage the information transferred from many-shot classes to augment the sparse training data for few-shot classes using a meta-generator.  ... 
arXiv:2008.09892v1 fatcat:tmmlddvj5vfsxcvbtihdxwmf4i

Data InStance Prior (DISP) in Generative Adversarial Networks [article]

Puneet Mangla, Nupur Kumari, Mayank Singh, Balaji Krishnamurthy, Vineeth N Balasubramanian
2021 arXiv   pre-print
Previous works have addressed training in low data setting by leveraging transfer learning and data augmentation techniques.  ...  We propose a novel transfer learning method for GANs in the limited data domain by leveraging informative data prior derived from self-supervised/supervised pre-trained networks trained on a diverse source  ...  We compare our approach with this method in the few-shot data setting.  ... 
arXiv:2012.04256v2 fatcat:hl4fepdoqvdfzbim4toyl27xd4

Bowtie Networks: Generative Modeling for Joint Few-Shot Recognition and Novel-View Synthesis [article]

Zhipeng Bao, Yu-Xiong Wang, Martial Hebert
2021 arXiv   pre-print
as ways of data augmentation, especially in the low-data regime.  ...  We propose a novel task of joint few-shot recognition and novel-view synthesis: given only one or few images of a novel object from arbitrary views with only category annotation, we aim to simultaneously  ...  Few-Shot Setting: The few-shot dataset consists of one or only a few images per category, which makes our problem even more challenging.  ... 
arXiv:2008.06981v2 fatcat:t5fxbkexkrbnpheidxlfznncoi

Skin Disease Analysis with Limited Data in particular Rosacea: A Review and Recommended Framework

Anwesha Mohanty, Alistair Sutherland, Marija Bezbradica, Hossein Javidnia.
2022 IEEE Access  
The challenges in data acquisition for a few lamentably neglected skin conditions such as rosacea are an excellent instance to explore the possibilities of improving computer-aided skin disease diagnosis  ...  With data scarcity in mind, the possible techniques explored and discussed include Generative Adversarial Networks, Meta-Learning, Few-Shot classification, and 3D face modelling.  ...  Considering the limitations of transfer learning and data augmentation, this study has considered a few methods such as GANs, 3D modelling and meta-learning with few-shot classification to improve medical  ... 
doi:10.1109/access.2022.3165574 fatcat:52576aerbbfizciwqeqtufg5ue

Generating Expensive Relationship Features from Cheap Objects

Xiaogang Wang, Qianru Sun, Tat-Seng Chua, Marcelo H. Ang Jr.
2019 British Machine Vision Conference  
can be consistently improved in various settings such as zero-shot and low-shot.  ...  Inspired by the data augmentation methods, we propose a novel Semantic Transform Generative Adversarial Network (ST-GAN) that synthesizes relationship features for rare objects, conditioned on the features  ...  NExT++ research which is supported by the National Research Foundation, Prime Minister's Office, Singapore under its IRC@SG Funding Initiative, and it is part of CREATE program, Singapore-MIT Alliance for  ... 
dblp:conf/bmvc/WangSCA19 fatcat:s3qeltenwza6lfeix4ehwkhwea

Leveraging Local Domains for Image-to-Image Translation [article]

Anthony Dell'Eva, Fabio Pizzati, Massimo Bertozzi, Raoul de Charette
2022 arXiv   pre-print
Relying on a simple geometrical guidance, we train a patch-based GAN on few source data and hallucinate a new unseen domain which subsequently eases transfer learning to target.  ...  In this paper, we leverage human knowledge about spatial domain characteristics which we refer to as 'local domains' and demonstrate its benefit for image-to-image translation.  ...  An interesting characteristic of our method is that it trains few shots on source only, leveraging only minimal human knowledge about the target.  ... 
arXiv:2109.04468v3 fatcat:vnh5rhjrvngibbcsl4gjd2esbi

Unsupervised Few-Shot Action Recognition via Action-Appearance Aligned Meta-Adaptation [article]

Jay Patravali, Gaurav Mittal, Ye Yu, Fuxin Li, Mei Chen
2021 arXiv   pre-print
We present MetaUVFS as the first Unsupervised Meta-learning algorithm for Video Few-Shot action recognition.  ...  MetaUVFS comprises a novel Action-Appearance Aligned Meta-adaptation (A3M) module that learns to focus on the action-oriented video features in relation to the appearance features via explicit few-shot  ...  Few-shot learning is highly relevant for videos because collecting large-scale labeled video data is extra challenging with the additional temporal dimension.  ... 
arXiv:2109.15317v2 fatcat:36fm5sd6azhzpdcfr2hoh3lyjm

F-VAEGAN-D2: A Feature Generating Framework for Any-Shot Learning

Yongqin Xian, Saurabh Sharma, Bernt Schiele, Zeynep Akata
2019 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
In this paper, we tackle any-shot learning problems i.e. zero-shot and few-shot, in a unified feature generating framework that operates in both inductive and transductive learning settings.  ...  When labeled training data is scarce, a promising data augmentation approach is to generate visual features of unknown classes using their attributes.  ...  In contrast to those inductive approaches that only use labeled data from seen classes, transductive zero-shot learning methods additionally leverage unlabeled data from unseen classes.  ... 
doi:10.1109/cvpr.2019.01052 dblp:conf/cvpr/XianSSA19 fatcat:ri2rdtqoqbc7vnwipzixyaqwoy

Few-shot 3D Multi-modal Medical Image Segmentation using Generative Adversarial Learning [article]

Arnab Kumar Mondal, Jose Dolz, Christian Desrosiers
2018 arXiv   pre-print
We address the problem of segmenting 3D multi-modal medical images in scenarios where very few labeled examples are available for training.  ...  Leveraging the recent success of adversarial learning for semi-supervised segmentation, we propose a novel method based on Generative Adversarial Networks (GANs) to train a segmentation model with both  ...  It also provides a new technique for few-shot learning, obviating the need for an initial pre-trained network by leveraging the semi-supervised learning ability of GANs.  ... 
arXiv:1810.12241v1 fatcat:qqlum7slw5hmvneyh7y5rqazwu
« Previous Showing results 1 — 15 out of 1,715 results