Filters








97 Hits in 3.6 sec

Region Rebalance for Long-Tailed Semantic Segmentation [article]

Jiequan Cui, Yuhui Yuan, Zhisheng Zhong, Zhuotao Tian, Han Hu, Stephen Lin, Jiaya Jia
2022 arXiv   pre-print
To verify the flexibility and effectiveness of our method, we apply the region rebalance module into various semantic segmentation methods, such as Deeplabv3+, OCRNet, and Swin.  ...  In this paper, we study the problem of class imbalance in semantic segmentation. We first investigate and identify the main challenges of addressing this issue through pixel rebalance.  ...  To address the long-tailed semantic segmentation problem, we first apply the well-known long-tailed image classification method to the semantic segmentation task.  ... 
arXiv:2204.01969v1 fatcat:udyf4zguzzehxaasnzkufvxg7a

Unsupervised Domain Adaptation for Semantic Segmentation by Content Transfer [article]

Suhyeon Lee, Junhyuk Hyun, Hongje Seong, Euntai Kim
2020 arXiv   pre-print
Even though we perfectly extract content for semantic segmentation in the real domain, another main challenge, the class imbalance problem, still exists in UDA for semantic segmentation.  ...  Here, only the content has cues for semantic segmentation, and the style makes the domain gap.  ...  We can train the segmentation model with the definite label for the transferred region.  ... 
arXiv:2012.12545v1 fatcat:qxdokvfw65ekto4cydmindcn2m

Locally Adaptive Learning Loss for Semantic Image Segmentation [article]

Jinjiang Guo, Pengyuan Ren, Aiguo Gu, Jian Xu, Weixin Wu
2018 arXiv   pre-print
We propose a novel locally adaptive learning estimator for enhancing the inter- and intra- discriminative capabilities of Deep Neural Networks, which can be used as improved loss layer for semantic image  ...  segmentation tasks.  ...  Moreover, most semantic segmentation datasets exhibit long tail distributions with few object categories, which means inter-and intra-classes are imbalanced, and consequently biasing networks training  ... 
arXiv:1802.08290v2 fatcat:aou5jrcjqjfdhk5n4q6fpsu5t4

FASA: Feature Augmentation and Sampling Adaptation for Long-Tailed Instance Segmentation [article]

Yuhang Zang, Chen Huang, Chen Change Loy
2021 arXiv   pre-print
Recent methods for long-tailed instance segmentation still struggle on rare object classes with few training data.  ...  We show FASA is a fast, generic method that can be easily plugged into standard or long-tailed segmentation frameworks, with consistent performance gains and little added cost.  ...  Surprisingly, data augmentation as a simple technique, has been barely studied for long-tailed instance segmentation.  ... 
arXiv:2102.12867v2 fatcat:rhwbruwc5zekrdclvpd34mxpoa

A Survey on Long-Tailed Visual Recognition

Lu Yang, He Jiang, Qing Song, Jun Guo
2022 International Journal of Computer Vision  
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.  ...  Finally, we provide several future directions for the development of long-tailed learning to provide more ideas for readers.  ...  For example, for the pixel-level semantic segmentation task, only few studies [187] have addressed the long-tailed phenomenon.  ... 
doi:10.1007/s11263-022-01622-8 fatcat:gtlfuw6igbd6xixqhynnqxubw4

MOSAIC: Mobile Segmentation via decoding Aggregated Information and encoded Context [article]

Weijun Wang, Andrew Howard
2021 arXiv   pre-print
We present a next-generation neural network architecture, MOSAIC, for efficient and accurate semantic image segmentation on mobile devices.  ...  MOSAIC is designed using commonly supported neural operations by diverse mobile hardware platforms for flexible deployment across various mobile platforms.  ...  Enet: A deep neural network architecture for dataset for semantic urban scene understanding. In Proc. real-time semantic segmentation.  ... 
arXiv:2112.11623v1 fatcat:ayhux3gbybgf5edd7m77ov5l7u

Hybrid Variability Aware Network (HVANet): A Self-Supervised Deep Framework for Label-Free SAR Image Change Detection

Jian Wang, Yinghua Wang, Hongwei Liu
2022 Remote Sensing  
More importantly, for bitemporal SAR images, changed regions tend to present highly variable sizes, irregular shapes, and different textures, typically referred to as hybrid variabilities, further bringing  ...  In this paper, we argue that these internal hybrid variabilities can also be used for learning stronger feature representation, and we propose a hybrid variability aware network (HVANet) for completely  ...  Acknowledgments: The authors would like to thank all reviewers and editors for their comments on this paper. Conflicts of Interest: The authors declare no conflict of interest.  ... 
doi:10.3390/rs14030734 fatcat:z6gojgx3lnbire6rqcnd32oh44

Towards Balanced Learning for Instance Recognition

Jiangmiao Pang, Kai Chen, Qi Li, Zhihai Xu, Huajun Feng, Jianping Shi, Wanli Ouyang, Dahua Lin
2021 International Journal of Computer Vision  
To mitigate the adverse effects caused thereby, we propose Libra R-CNN, a simple yet effective framework towards balanced learning for instance recognition.  ...  It integrates IoU-balanced sampling, balanced feature pyramid, and objective re-weighting, respectively for reducing the imbalance at sample, feature, and objective level.  ...  These problems will become even worse when transferring to the long-tailed cases, which will be discussed below.  ... 
doi:10.1007/s11263-021-01434-2 fatcat:lvblzs5cnzfz7daf2exgx6u3pq

Bounding Boxes Are All We Need: Street View Image Classification via Context Encoding of Detected Buildings [article]

Kun Zhao, Yongkun Liu, Siyuan Hao, Shaoxing Lu, Hongbin Liu, Lijian Zhou
2020 arXiv   pre-print
The dataset can be used not only for street view image classification, but also for multi-class building detection.  ...  ACKNOWLEDGMENT The authors would like to thank the authors of reference [12] for publishing the BIC GSV dataset including city scale GSV images.  ...  Thanks to those who participated in manual annotation for building detection: Yu Ma, Shanshan Lin, Ying Guo and Kaixin Li, and who participated in manual  ... 
arXiv:2010.01305v2 fatcat:cpf3nfh3fnampfl72vxbo2cy5y

MICA: A Holistic Approach to Fast In-Memory Key-Value Storage

Hyeontaek Lim, Dongsu Han, David G. Andersen, Michael Kaminsky
2014 Symposium on Networked Systems Design and Implementation  
We would like to thank Nick Feamster, John Ousterhout, Dong Zhou, Yandong Mao, Wyatt Lloyd, and our NSDI reviewers for their valuable feedback, and Prabal Dutta for shepherding this paper.  ...  Acknowledgments This work was supported by funding from the National Science Foundation under awards CCF-0964474 and CNS-1040801, Intel via the Intel Science and Technology Center for Cloud Computing (  ...  When an item is deleted, MICA coalesces any adjacent free regions using boundary tags [26] to recreate a large free region.  ... 
dblp:conf/nsdi/LimHAK14 fatcat:jsilhrm3azh7dmums7frq4fqeq

Rebalanced Siamese Contrastive Mining for Long-Tailed Recognition [article]

Zhisheng Zhong, Jiequan Cui, Zeming Li, Eric Lo, Jian Sun, Jiaya Jia
2022 arXiv   pre-print
Extensive experiments demonstrate that ResCom outperforms the previous methods by large margins on multiple long-tailed recognition benchmarks.  ...  mathematical analysis and simulation results, we claim that supervised contrastive learning suffers a dual class-imbalance problem at both the original batch and Siamese batch levels, which is more serious than long-tailed  ...  segmentation [6] .  ... 
arXiv:2203.11506v2 fatcat:yegn6pzzhfeydjyr6mpub4crwa

Shenango: Achieving High CPU Efficiency for Latency-sensitive Datacenter Workloads

Amy Ousterhout, Joshua Fried, Jonathan Behrens, Adam Belay, Hari Balakrishnan
2019 Symposium on Networked Systems Design and Implementation  
We thank Henry Qin for helping us evaluate Arachne. Amy Ousterhout was supported by an NSF Fellowship and a Hertz Foundation Fellowship.  ...  Acknowledgments We thank our shepherd KyoungSoo Park, the anonymous reviewers, John Ousterhout, Tom Anderson, Frans Kaashoek, Nickolai Zeldovich, and other members of PDOS for their useful feedback.  ...  Similarly, Linux rebalances tasks across cores primarily in response to millisecond-scale timer ticks.  ... 
dblp:conf/nsdi/OusterhoutFBBB19 fatcat:vdng2w4l35elrkjidc6e6ht6tq

Music Information Retrieval: Recent Developments and Applications

Markus Schedl, Emilia Gómez, Julián Urbano
2014 Foundations and Trends in Information Retrieval  
These in turn enable a wide variety of music retrieval tasks, such as semantic music search or music identification ("query by example").  ...  We first elaborate on well-established and proven methods for feature extraction and music indexing, from both the audio signal and contextual data sources about music items, such as web pages or collaborative  ...  Bosch for his comments on the manuscript. Furthermore, the authors would like to express their gratitude to the anonymous reviewers for their highly valuable suggestions for improving the manuscript.  ... 
doi:10.1561/1500000042 fatcat:c5tjdcy3xrfqvp6isnktbr6lpy

Monocular Depth Estimation with Joint Attention Feature Distillation and Wavelet-Based Loss Function

Peng Liu, Zonghua Zhang, Zhaozong Meng, Nan Gao
2020 Sensors  
This paper proposes a new deep convolutional neural network for monocular depth estimation.  ...  Second, we adopted a wavelet-based loss function for network training, which improves loss function effectiveness by obtaining more structural details.  ...  [36] proposed a 3D representation for semantic segmentation and depth estimation from a single image. Lin et al.  ... 
doi:10.3390/s21010054 pmid:33374278 fatcat:gokfbojwsvdxdam4n3ffptvr6i

Deep Imbalanced Learning for Face Recognition and Attribute Prediction [article]

Chen Huang, Yining Li, Chen Change Loy, Xiaoou Tang
2019 arXiv   pre-print
In this paper, we conduct extensive and systematic experiments to validate the effectiveness of these classic schemes for representation learning on class-imbalanced data.  ...  Data for face analysis often exhibit highly-skewed class distribution, i.e., most data belong to a few majority classes, while the minority classes only contain a scarce amount of instances.  ...  [55] further combined a deep semantic segmentation network to guide attribute prediction to the corresponding local region.  ... 
arXiv:1806.00194v2 fatcat:3qbovwkh7bfknaudlcjclui4gi
« Previous Showing results 1 — 15 out of 97 results