Filters








19,441 Hits in 6.0 sec

REZCR: Zero-shot Character Recognition via Radical Extraction [article]

Xiaolei Diao, Daqian Shi, Hao Tang, Yanzeng Li, Lei Wu, Hao Xu
2022 arXiv   pre-print
In this paper, we propose a zero-shot character recognition framework via radical extraction (REZCR) to improve the recognition performance of few-sample character categories in the tail.  ...  The RIE aims to recognize candidate radicals and their possible structural relations from character images.  ...  According to the above-mentioned discussions, we propose a novel method for zero-shot character recognition via radical extraction, namely REZCR.  ... 
arXiv:2207.05842v2 fatcat:fiueuf4mava3rbfbbrlla2luem

Zero-Shot Visual Recognition via Semantic Attention-based Compare Network

Fudong Nian, Yikun Sheng, Junfeng Wang, Teng Li
2020 IEEE Access  
There is an increasing interest in machine learning area for achieving this humans ability via computer, which is named zero-shot visual recognition.  ...  The goal of zero-shot visual recognition is to predict the class label [38] and [39] , almost all state-of-theart zero-shot visual recognition methods can be unified into a mapping-based framework.  ... 
doi:10.1109/access.2020.2971174 fatcat:4ek3j2f4jva4bnn3cl5yt2szwq

Zero-Shot Visual Recognition via Bidirectional Latent Embedding [article]

Qian Wang, Ke Chen
2017 arXiv   pre-print
Unlike most of the existing zero-shot visual recognition methods, we propose a stagewise bidirectional latent embedding framework to two subsequent learning stages for zero-shot visual recognition.  ...  Zero-shot learning for visual recognition, e.g., object and action recognition, has recently attracted a lot of attention.  ...  Also the authors are grateful to Ziming Zhang at Boston University for providing their source code in structured prediction and Yongqin Xian at Max Planck Institute for Informatics for providing their  ... 
arXiv:1607.02104v4 fatcat:35tawiiuwvgjxhkolsehwstkte

DeepStruct: Pretraining of Language Models for Structure Prediction [article]

Chenguang Wang, Xiao Liu, Zui Chen, Haoyun Hong, Jie Tang, Dawn Song
2022 arXiv   pre-print
Our structure pretraining enables zero-shot transfer of the learned knowledge that models have about the structure tasks.  ...  We study the performance of this approach on 28 datasets, spanning 10 structure prediction tasks including open information extraction, joint entity and relation extraction, named entity recognition, relation  ...  We enable the zero-shot transfer to structure prediction tasks via converting the downstream tasks to one or a combination of the pretraining tasks.  ... 
arXiv:2205.10475v1 fatcat:jjvsqdrhvzdyvllcob53mkyika

Zero-Shot Visual Recognition via Bidirectional Latent Embedding

Qian Wang, Ke Chen
2017 International Journal of Computer Vision  
Unlike most of the existing zero-shot visual recognition methods, we propose a stagewise bidirectional latent embedding framework to two subsequent learning stages for zero-shot visual recognition.  ...  Zero-shot learning for visual recognition, e.g., object and action recognition, has recently attracted a lot of attention.  ...  Also the authors are grateful to Ziming Zhang at Boston University for providing their source code in structured prediction and Yongqin Xian at Max Planck Institute for Informatics for providing their  ... 
doi:10.1007/s11263-017-1027-5 fatcat:ofepocl3gvgzbn4y6dwgazmgh4

Transfer Learning in a Transductive Setting

Marcus Rohrbach, Sandra Ebert, Bernt Schiele
2013 Neural Information Processing Systems  
More specifically we adapt a graph-based learning algorithm -so far only used for semi-supervised learningto zero-shot and few-shot learning.  ...  Second, we exploit the manifold structure of novel classes.  ...  We evaluated this approach on three diverse datasets for image and video-activity recognition, consistently improving performance over the state-of-the-art for zero-shot and few-shot prediction.  ... 
dblp:conf/nips/RohrbachES13 fatcat:b3mufpaek5dqnfqogtaqbf3jzm

Zero-shot Recognition via Semantic Embeddings and Knowledge Graphs [article]

Xiaolong Wang, Yufei Ye, Abhinav Gupta
2018 arXiv   pre-print
We consider the problem of zero-shot recognition: learning a visual classifier for a category with zero training examples, just using the word embedding of the category and its relationship to other categories  ...  At test time, these filters are used to predict the visual classifiers of unseen categories. We show that our approach is robust to noise in the KG.  ...  By using the predicted embedding to perform nearest neighbor search, DeViSE scales up the zero-shot recognition to thousands of classes.  ... 
arXiv:1803.08035v2 fatcat:fbe2t3l2kfeexnm7oggb4v4hqe

Zero-Shot Recognition via Semantic Embeddings and Knowledge Graphs

Xiaolong Wang, Yufei Ye, Abhinav Gupta
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
We consider the problem of zero-shot recognition: learning a visual classifier for a category with zero training examples, just using the word embedding of the category and its relationship to other categories  ...  At test time, these filters are used to predict the visual classifiers of unseen categories. We show that our approach is robust to noise in the KG.  ...  By using the predicted embedding to perform nearest neighbor search, DeViSE scales up the zero-shot recognition to thousands of classes.  ... 
doi:10.1109/cvpr.2018.00717 dblp:conf/cvpr/0004YG18 fatcat:uh3n5xgf3fc7xn7uxufyn74ipu

Hierarchical Prototype Learning for Zero-Shot Recognition [article]

Xingxing Zhang, Shupeng Gui, Zhenfeng Zhu, Yao Zhao, Ji Liu
2019 arXiv   pre-print
In this paper, we propose a hierarchical prototype learning formulation to provide a systematical solution (named HPL) for zero-shot recognition.  ...  Zero-Shot Learning (ZSL) has received extensive attention and successes in recent years especially in areas of fine-grained object recognition, retrieval, and image captioning.  ...  The goal of zero-shot recognition is to predict the labels of test samples in D u by learning a classifier f u (·) : X u → U.  ... 
arXiv:1910.11671v2 fatcat:pwtkp2dxvrhrvg7lhmggulayqi

Transductive Multi-class and Multi-label Zero-shot Learning [article]

Yanwei Fu, Yongxin Yang, Timothy M. Hospedales, Tao Xiang, Shaogang Gong
2015 arXiv   pre-print
Recently, zero-shot learning (ZSL) has received increasing interest.  ...  semantic representation is learned from the auxiliary dataset by either classification or regression models and applied directly to map each instance into the same semantic representation space where a zero-shot  ...  With this synthetic dataset, we are able to propose two new multi-label algorithms -direct multi-label zero-shot prediction (DMP) and transductive multi-label zero-shot prediction (TraMP).  ... 
arXiv:1503.07884v1 fatcat:or4zahtjj5atpoxks5n4dilega

PointCLIP: Point Cloud Understanding by CLIP [article]

Renrui Zhang, Ziyu Guo, Wei Zhang, Kunchang Li, Xupeng Miao, Bin Cui, Yu Qiao, Peng Gao, Hongsheng Li
2021 arXiv   pre-print
Recently, zero-shot and few-shot learning via Contrastive Vision-Language Pre-training (CLIP) have shown inspirational performance on 2D visual recognition, which learns to match images with their corresponding  ...  Specifically, we encode a point cloud by projecting it into multi-view depth maps without rendering, and aggregate the view-wise zero-shot prediction to achieve knowledge transfer from 2D to 3D.  ...  Related Work Zero-shot Learning in 3D. The objective of zero-shot learning is to enable recognition of "unseen" objects which are not adopted during training.  ... 
arXiv:2112.02413v1 fatcat:hjiexbelk5djvatobzqnk4zsqq

Unified BERT for Few-shot Natural Language Understanding [article]

Junyu Lu, Ping Yang, Ruyi Gan, Jing Yang, Jiaxing Zhang
2022 arXiv   pre-print
By using the biaffine to model scores pair of the start and end position of the original text, various classification and extraction structures can be converted into a universal, span-decoding approach  ...  Experiments show that UBERT wins the first price in the 2022 AIWIN - World Artificial Intelligence Innovation Competition, Chinese insurance few-shot multi-task track, and realizes the unification of extensive  ...  Due to the consistency objectives of downstream tasks and pre-training, these pre-trained LMs perform well on few-shot and zero-shot scenario.  ... 
arXiv:2206.12094v2 fatcat:hyax4rzmhfgwfjvmzs5zrkky3e

Generalized Zero-Shot Learning for Action Recognition with Web-Scale Video Data [article]

Kun Liu, Wu Liu, Huadong Ma, Wenbing Huang, Xiongxiong Dong
2017 arXiv   pre-print
To our best knowledge, this is the first work to study video action recognition under the generalized zero-shot setting.  ...  Action recognition in surveillance video makes our life safer by detecting the criminal events or predicting violent emergencies. However, efficient action recognition is not free of difficulty.  ...  However, it is likely to violate the zero-shot learning assumption since extracting video feature via a pre-trained C3D on a large dataset [37] that contains zero-shot testing classes.  ... 
arXiv:1710.07455v1 fatcat:datwl63c5jd2hiylkz7636lra4

Disentangled Action Recognition with Knowledge Bases [article]

Zhekun Luo, Shalini Ghosh, Devin Guillory, Keizo Kato, Trevor Darrell, Huijuan Xu
2022 arXiv   pre-print
DARK trains a factorized model by first extracting disentangled feature representations for verbs and nouns, and then predicting classification weights using relations in external knowledge graphs.  ...  To address this issue, we propose our approach: Disentangled Action Recognition with Knowledge-bases (DARK), which leverages the inherent compositionality of actions.  ...  This motivates us to study the problem of zero-shot compositional action recognition, which aims to predict action with components beyond the vocabularies in train data.  ... 
arXiv:2207.01708v1 fatcat:qywim6ipxvasvlwqqu7atfugny

Learning Structured Semantic Embeddings for Visual Recognition [article]

Dong Li, Hsin-Ying Lee, Jia-Bin Huang, Shengjin Wang, Ming-Hsuan Yang
2017 arXiv   pre-print
Extensive evaluations demonstrate the effectiveness of the proposed structured embeddings for single-label classification, multi-label classification, and zero-shot recognition.  ...  Numerous embedding models have been recently explored to incorporate semantic knowledge into visual recognition.  ...  , and zero-shot recognition.  ... 
arXiv:1706.01237v1 fatcat:fdsvdt2ijjawpludyjyfjt6ucy
« Previous Showing results 1 — 15 out of 19,441 results