11,040 Hits in 7.0 sec

Efficient Annotation and Learning for 3D Hand Pose Estimation: A Survey [article]

Takehiko Ohkawa and Ryosuke Furuta and Yoichi Sato
2022 arXiv   pre-print
Since these annotation methods are not always available on a large scale, we examined methods of learning 3D hand poses when we do not have enough annotated data, namely self-supervised pre-training, semi-supervised  ...  In particular, we study recent approaches for 3D hand pose annotation and learning methods with limited annotated data.  ...  On the contrary, GRAB [55] was built with a motion capture system for human hands and body, but it does not possess visual modality.  ... 
arXiv:2206.02257v2 fatcat:6qaf2sieaffurl3z7bfdxpwcu4

A Probabilistic Semi-Supervised Approach to Multi-Task Human Activity Modeling [article]

Judith Bütepage, Hedvig Kjellström, Danica Kragic
2019 arXiv   pre-print
We focus here on the tasks of action classification, detection, prediction and anticipation as well as motion prediction and synthesis based on 3D human activity data recorded with Kinect.  ...  In this paper we present a semi-supervised probabilistic deep latent variable model that can represent both discrete labels and continuous observations as well as latent dynamics over time.  ...  The H36M dataset consists of motion capture data and is often used for human motion prediction experiments.  ... 
arXiv:1809.08875v3 fatcat:dvgoa2kh2bcw7iet5xqugnc3ga

Improved Generalization of Heading Direction Estimation for Aerial Filming Using Semi-supervised Regression [article]

Wenshan Wang, Aayush Ahuja, Yanfu Zhang, Rogerio Bonatti, Sebastian Scherer
2019 arXiv   pre-print
Towards improving generalization with less amount of labeled data, this paper presents a semi-supervised algorithm for heading direction estimation problem.  ...  We show that by leveraging unlabeled sequences, the amount of labeled data required can be significantly reduced.  ...  We compared our semi-supervised method with supervised one using different number of labeled data and the result is shown in Figure 5 .  ... 
arXiv:1903.11174v1 fatcat:3i477cojyrcblpypuzl5krfg7u

Joint-bone Fusion Graph Convolutional Network for Semi-supervised Skeleton Action Recognition [article]

Zhigang Tu, Jiaxu Zhang, Hongyan Li, Yujin Chen, Junsong Yuan
2022 arXiv   pre-print
the latent functional correlation between joints and bones for action recognition. 2) Most of these works are performed in the supervised learning way, which heavily relies on massive labeled training  ...  NTU-RGB+D and Kinetics-Skeleton, demonstrate that our model achieves the state-of-the-art performance for semi-supervised skeleton-based action recognition and is also useful for fully-supervised methods  ...  It was also supported by the Central University Basic Research Fund of China (No.2042020KF0016).  ... 
arXiv:2202.04075v1 fatcat:g6xs3q24yvdsxkecdwvum6lgnm

Special issue on "visual semantic analysis with weak supervision"

Luming Zhang, Yang Yang, Rongrong Ji, Roger Zimmermann
2017 Multimedia Systems  
1 National University of Singapore, Singapore, Singapore Acknowledgments We also thank the reviewers for their efforts to guarantee the high quality of this special issue.  ...  In "A Human Motion Feature based on Semi-supervised Learning of GMM", Qi et al. presented a novel statistic feature to represent each motion according to the pre-labeled categories of key-poses.  ...  Compared with the laborintensive labeling in fully supervised setting, the transferring mechanism can greatly reduce the human effort, especially given the tremendous visual data on the Websites.  ... 
doi:10.1007/s00530-016-0527-4 fatcat:72hcjiiwfzbk7mdrsumuia7rzy

Author Index

2010 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition  
Object and Human Pose in Human-Object Interaction Activities Connecting Modalities: Semi-supervised Segmentation and Annotation of Images Using Unaligned Text Corpora Felzenszwalb, Pedro F.  ...  Boundary Learning by Optimization with Topological Constraints Helten, Thomas Multisensor-Fusion for 3D Full-Body Human Motion Capture Hendriks, Emile Workshop: Capturing Appearance Variation in  ... 
doi:10.1109/cvpr.2010.5539913 fatcat:y6m5knstrzfyfin6jzusc42p54

A human motion feature based on semi-supervised learning of GMM

Tian Qi, Yinfu Feng, Jun Xiao, Hanzhi Zhang, Yueting Zhuang, Xiaosong Yang, Jianjun Zhang
2014 Multimedia Systems  
A probabilistic model is trained with semi-supervised learning of the Gaussian mixture model (GMM).  ...  A new category Abstract Using motion capture to create naturally looking motion sequences for virtual character animation has become a standard procedure in the games and visual effects industry.  ...  Introduction The growing popularity of motion capture (mocap) technique in feature films and interactive entertainments has led to an explosive growth of motion data.  ... 
doi:10.1007/s00530-014-0429-2 fatcat:xgjk5prdpvfgjjx7twbba7bazy

Learning to Localize Sound Sources in Visual Scenes: Analysis and Applications [article]

Arda Senocak, Tae-Hyun Oh, Junsik Kim, Ming-Hsuan Yang, In So Kweon
2019 arXiv   pre-print
We show that the false conclusions can be effectively corrected even with a small amount of supervision, i.e., semi-supervised setup.  ...  To fix this issue, we extend our network to the supervised and semi-supervised network settings via a simple modification due to the general architecture of our two-stream network.  ...  Despite using less supervised data, the semi-supervised approach also gives comparably accurate localization results. supervision.  ... 
arXiv:1911.09649v1 fatcat:gcan5noupzdkhkooidx73n3lqu

Semi-Supervised Haptic Material Recognition for Robots using Generative Adversarial Networks [article]

Zackory Erickson, Sonia Chernova, Charles C. Kemp
2017 arXiv   pre-print
However, collecting labeled training data with a robot is often more difficult than unlabeled data.  ...  We present a semi-supervised learning approach for material recognition that uses generative adversarial networks (GANs) with haptic features such as force, temperature, and vibration.  ...  We thank the developers of Keras, which we used throughout our experiments [26] .  ... 
arXiv:1707.02796v2 fatcat:m7nnnfwm35autj5m6bvdfjmna4

SMATE: Semi-Supervised Spatio-Temporal Representation Learning on Multivariate Time Series [article]

Jingwei Zuo, Karine Zeitouni, Yehia Taher
2021 arXiv   pre-print
We compare it with 13 state-of-the-art baseline methods for fully supervised tasks and four baselines for semi-supervised tasks. The results show the reliability and efficiency of our proposed method.  ...  In particular, label shortage is a real challenge for the classification task on MTS, considering its complex dimensional and sequential data structure.  ...  ACKNOWLEDGEMENTS This research was supported by DATAIA convergence institute as part of the Programme d'Investissement d'Avenir (ANR-17-CONV-0003) operated by DAVID Lab, UVSQ, Université Paris-Saclay.  ... 
arXiv:2110.00578v2 fatcat:lvplo2auovdfzka66yaghsvi5q

Predicting the What and How - a Probabilistic Semi-Supervised Approach to Multi-Task Human Activity Modeling

Judith Butepage, Hedvig Kjellstrom, Danica Kragic
2019 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)  
The aim of this work is to encourage research within the field of human activity modeling based on mixed categorical and continuous data.  ...  Video-based prediction of human activity is usually performed on one of two levels: either a model is trained to anticipate high-level action labels or it is trained to predict future trajectories either  ...  The semi-supervised VAE [5] can handle labeled and unlabeled data.  ... 
doi:10.1109/cvprw.2019.00352 dblp:conf/cvpr/ButepageKK19 fatcat:42zh4kx5wvgmrb7d5bffnzoxli

Multiview Supervision By Registration [article]

Yilun Zhang, Hyun Soo Park
2019 arXiv   pre-print
This paper presents a semi-supervised learning framework to train a keypoint detector using multiview image streams given the limited labeled data (typically <4%).  ...  We leverage the complementary relationship between multiview geometry and visual tracking to provide three types of supervisionary signals to utilize the unlabeled data: (1) keypoint detection in one view  ...  The resulting network precisely detects the keypoints of both non-human and human subjects with highly limited labeled data (< 4%).  ... 
arXiv:1811.11251v2 fatcat:3fs35r2k3nhktcivizi2dnpqyy

Learning to Localize Sound Source in Visual Scenes [article]

Arda Senocak, Tae-Hyun Oh, Junsik Kim, Ming-Hsuan Yang, In So Kweon
2018 arXiv   pre-print
We show that even with a few supervision, false conclusion is able to be corrected and the source of sound in a visual scene can be localized effectively.  ...  Moreover, although our network is formulated within the unsupervised learning framework, it can be extended to a unified architecture with a simple modification for the supervised and semi-supervised learning  ...  We plot the success rate of the test samples according to cIoU threshold in Figure 9 . We measure the effect of the number of labeled samples in the semi-supervised scenario in Table 2 .  ... 
arXiv:1803.03849v1 fatcat:vubtdsrybnayxizc4wxwvd2kdi

Semi-Supervised 3D Hand-Object Poses Estimation with Interactions in Time [article]

Shaowei Liu, Hanwen Jiang, Jiarui Xu, Sifei Liu, Xiaolong Wang
2021 arXiv   pre-print
Going beyond limited 3D annotations in a single image, we leverage the spatial-temporal consistency in large-scale hand-object videos as a constraint for generating pseudo labels in semi-supervised learning  ...  Estimating 3D hand and object pose from a single image is an extremely challenging problem: hands and objects are often self-occluded during interactions, and the 3D annotations are scarce as even humans  ...  Despite the efforts, current approaches still highly rely on human annotations for 3D poses, which are extremely difficult to obtain: Researchers have been collecting data with motion capture [61, 17,  ... 
arXiv:2106.05266v1 fatcat:ig3f2sw4z5ezlfpipw77y5i5tm

Self-supervised Transfer Learning for Instance Segmentation through Physical Interaction [article]

Andreas Eitel and Nico Hauff and Wolfram Burgard
2020 arXiv   pre-print
To overcome the time-consuming process of manually labeling data for new environments, we present a transfer learning approach for robots that learn to segment objects by interacting with their environment  ...  in a self-supervised manner.  ...  [33] present semi-automatic data labeling for semantic segmentation. Aforementioned methods are self-supervised but not interactive. More recently, Danielczuk et al.  ... 
arXiv:2005.09484v1 fatcat:navvvb7im5h5xklbs6kynpbrga
« Previous Showing results 1 — 15 out of 11,040 results