Filters








382 Hits in 3.6 sec

Balanced Meta-Softmax for Long-Tailed Visual Recognition [article]

Jiawei Ren, Cunjun Yu, Shunan Sheng, Xiao Ma, Haiyu Zhao, Shuai Yi, Hongsheng Li
2020 arXiv   pre-print
In our experiments, we demonstrate that Balanced Meta-Softmax outperforms state-of-the-art long-tailed classification solutions on both visual recognition and instance segmentation tasks.  ...  In addition, we introduce Balanced Meta-Softmax, applying a complementary Meta Sampler to estimate the optimal class sample rate and further improve long-tailed learning.  ...  We introduce Balanced Meta-Softmax (BALMS) for long-tailed visual recognition.  ... 
arXiv:2007.10740v3 fatcat:atpjzlqdvreopbfs2xod63ni3u

Deep Long-Tailed Learning: A Survey [article]

Yifan Zhang, Bingyi Kang, Bryan Hooi, Shuicheng Yan, Jiashi Feng
2021 arXiv   pre-print
Deep long-tailed learning, one of the most challenging problems in visual recognition, aims to train well-performing deep models from a large number of images that follow a long-tailed class distribution  ...  However, long-tailed class imbalance, a common problem in practical visual recognition tasks, often limits the practicality of deep network based recognition models in real-world applications, since they  ...  Balanced meta-softmax [86] developed a meta learning based sampling method to estimate the optimal sampling rates of different classes for long-tailed learning.  ... 
arXiv:2110.04596v1 fatcat:lpvt2x6cv5crxm2qxdctjrlkqq

Large-Scale Long-Tailed Recognition in an Open World

Ziwei Liu, Zhongqi Miao, Xiaohang Zhan, Jiayun Wang, Boqing Gong, Stella X. Yu
2019 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
We define Open Long-Tailed Recognition (OLTR) as learning from such naturally distributed data and optimizing the classification accuracy over a balanced test set which include head, tail, and open classes  ...  Real world data often have a long-tailed and open-ended distribution.  ...  Introduction Our visual world is inherently long-tailed and openended: The frequency distribution of visual categories in our daily life is long-tailed [38] , with a few common classes and many more rare  ... 
doi:10.1109/cvpr.2019.00264 dblp:conf/cvpr/0002MZWGY19 fatcat:rvtrbxlzuna7fkjciktazhovum

Domain Balancing: Face Recognition on Long-Tailed Domains

Dong Cao, Xiangyu Zhu, Xingyu Huang, Jianzhu Guo, Zhen Lei
2020 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
Long-tailed problem has been an important topic in face recognition task. However, existing methods only concentrate on the long-tailed distribution of classes.  ...  Differently, we devote to the long-tailed domain distribution problem, which refers to the fact that a small number of domains frequently appear while other domains far less existing.  ...  Particularly, the margin-based methods attain better results than the simple softmax loss for face recognition.  ... 
doi:10.1109/cvpr42600.2020.00571 dblp:conf/cvpr/CaoZHGL20 fatcat:l33anueej5gb3cr655ujnepmey

Domain Balancing: Face Recognition on Long-Tailed Domains [article]

Dong Cao, Xiangyu Zhu, Xingyu Huang, Jianzhu Guo, Zhen Lei
2020 arXiv   pre-print
Long-tailed problem has been an important topic in face recognition task. However, existing methods only concentrate on the long-tailed distribution of classes.  ...  Differently, we devote to the long-tailed domain distribution problem, which refers to the fact that a small number of domains frequently appear while other domains far less existing.  ...  Particularly, the margin-based methods attain better results than the simple softmax loss for face recognition.  ... 
arXiv:2003.13791v1 fatcat:lrdb4rxvwnhq5bq5nzchz7aluu

A Survey on Long-Tailed Visual Recognition

Lu Yang, He Jiang, Qing Song, Jun Guo
2022 International Journal of Computer Vision  
In this survey, we focus on the problems caused by long-tailed data distribution, sort out the representative long-tailed visual recognition datasets and summarize some mainstream long-tailed studies.  ...  Besides, we have studied four quantitative metrics for evaluating the imbalance, and suggest using the Gini coefficient to evaluate the long-tailedness of a dataset.  ...  This provides guidance for long-tailed visual recognition, that is, different solutions are adopted through the long-tailedness of data.  ... 
doi:10.1007/s11263-022-01622-8 fatcat:gtlfuw6igbd6xixqhynnqxubw4

Meta Feature Modulator for Long-tailed Recognition [article]

Renzhen Wang, Kaiqin Hu, Yanwen Zhu, Jun Shu, Qian Zhao, Deyu Meng
2020 arXiv   pre-print
To address this issue, we propose meta feature modulator (MFM), a meta-learning framework to model the difference between the long-tailed training data and the balanced meta data from the perspective of  ...  Extensive experiments on benchmark vision datasets substantiate the superiority of our approach on long-tailed recognition tasks beyond other state-of-the-art methods.  ...  Experimental for long-tailed scene recognition Indoor scene recognition is a challenging problem where visual phenomena naturally follows a skewed distribution.  ... 
arXiv:2008.03428v1 fatcat:6yeebcjs7rc7nkwsesvn4mtep4

Large-Scale Long-Tailed Recognition in an Open World [article]

Ziwei Liu, Zhongqi Miao, Xiaohang Zhan, Jiayun Wang, Boqing Gong, Stella X. Yu
2019 arXiv   pre-print
We define Open Long-Tailed Recognition (OLTR) as learning from such naturally distributed data and optimizing the classification accuracy over a balanced test set which include head, tail, and open classes  ...  Real world data often have a long-tailed and open-ended distribution.  ...  Open long-tail recognition optimizes for the overall accuracy of all categories while fairness analysis optimizes for several attribute-wise criteria.  ... 
arXiv:1904.05160v2 fatcat:rz4nt5kacbhgnhzjsecs6havii

Learning of Visual Relations: The Devil is in the Tails [article]

Alakh Desai, Tz-Ying Wu, Subarna Tripathi, Nuno Vasconcelos
2021 arXiv   pre-print
To test this hypothesis, we devise a new approach for training visual relationships models, which is inspired by state-of-the-art long-tailed recognition literature.  ...  DT2 employs a novel sampling approach, Alternating Class-Balanced Sampling (ACBS), to capture the interplay between the long-tailed entity and predicate distributions of visual relations.  ...  Sampling for visual relationships Similar to long-tailed object recognition, it is sensible to train a model for visual relations in two stages.  ... 
arXiv:2108.09668v1 fatcat:4vhp4x2ombevbjsm347ccdhx2i

BatchFormer: Learning to Explore Sample Relationships for Robust Representation Learning [article]

Zhi Hou, Baosheng Yu, Dacheng Tao
2022 arXiv   pre-print
By doing this, the proposed method enables the collaboration of different samples, e.g., the head-class samples can also contribute to the learning of the tail classes for long-tailed recognition.  ...  perform extensive experiments on over ten datasets and the proposed method achieves significant improvements on different data scarcity applications without any bells and whistles, including the tasks of long-tailed  ...  Many knowledge transfer approaches have been developed for long-tailed recognition via meta learning [73] , memory features [45, 88] , and virtual data generation [26, 30, 69] .  ... 
arXiv:2203.01522v2 fatcat:2nra2zxp55ch7kizekdrtvtije

Long-Tail Hashing

Yong Chen, Yuqing Hou, Shu Leng, Qing Zhang, Zhouchen Lin, Dell Zhang
2021 Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval  
exhibit a long-tail distribution.  ...  A critical part of LTHNet is its dynamic meta-embedding module extended with a determinantal point process which can adaptively realize visual knowledge transfer between head and tail classes, and thus  ...  Ablation Study Is dynamic meta-embedding (Section 4.2) indeed useful for long-tail hashing?  ... 
doi:10.1145/3404835.3462888 fatcat:rddcc2n5wbajba3oo547hb7vxu

Explored An Effective Methodology for Fine-Grained Snake Recognition [article]

Yong Huang, Aderon Huang, Wei Zhu, Yanming Fang, Jinghua Feng
2022 arXiv   pre-print
Secondly, we provide new loss functions to solve the long tail distribution with dataset.  ...  Fine-Grained Visual Classification (FGVC) is a longstanding and fundamental problem in computer vision and pattern recognition, and underpins a diverse set of real-world applications.  ...  We also thank Dr Wei Zhu, Dr Yanming Fang for their guidance.  ... 
arXiv:2207.11637v1 fatcat:dw4ddndjbbdnliu5j2hh6kibna

Balanced Softmax Cross-Entropy for Incremental Learning [article]

Quentin Jodelet, Xin Liu, Tsuyoshi Murata
2021 arXiv   pre-print
To address this problem, we propose the use of the Balanced Softmax Cross-Entropy loss and show that it can be combined with exiting methods for incremental learning to improve their performances while  ...  By using a small memory for rehearsal and knowledge distillation, recent methods have proven to be effective to mitigate catastrophic forgetting.  ...  [27] to address the label distribution shift between the training and testing in Long-Tailed Visual Recognition.  ... 
arXiv:2103.12532v2 fatcat:hchmilbxtnhb3m5oywka44eptq

Distilling Virtual Examples for Long-tailed Recognition [article]

Yin-Yin He, Jianxin Wu, Xiu-Shen Wei
2021 arXiv   pre-print
We tackle the long-tailed visual recognition problem from the knowledge distillation perspective by proposing a Distill the Virtual Examples (DiVE) method.  ...  Furthermore, additional analyses and experiments verify the virtual example interpretation, and demonstrate the effectiveness of tailored designs in DiVE for long-tailed problems.  ...  But, in long-tailed recognition, Fig. 3 Hence, we easily obtain the following conclusion: in order to obtain a balanced model for a long-tailed task, the virtual example distribution must be much flatter  ... 
arXiv:2103.15042v3 fatcat:zqmpgnklujgn7jn4bdq6owaay4

GistNet: a Geometric Structure Transfer Network for Long-Tailed Recognition [article]

Bo Liu, Haoxiang Li, Hao Kang, Gang Hua, Nuno Vasconcelos
2021 arXiv   pre-print
The problem of long-tailed recognition, where the number of examples per class is highly unbalanced, is considered.  ...  Experiments on two popular long-tailed recognition datasets show that GistNet outperforms existing solutions to this problem.  ...  This is especially true for fewshot classes. Visualization Conclusion This work addressed the long-tailed recognition problem.  ... 
arXiv:2105.00131v1 fatcat:rgkklcmtkrgz5bhhlw6oy3rfzm
« Previous Showing results 1 — 15 out of 382 results