Filters








2,477 Hits in 8.5 sec

Inconsistent Few-Shot Relation Classification via Cross-Attentional Prototype Networks with Contrastive Learning [article]

Hongru Wang, Zhijing Jin, Jiarun Cao, Gabriel Pui Cheong Fung, Kam-Fai Wong
2021 arXiv   pre-print
Standard few-shot relation classification (RC) is designed to learn a robust classifier with only few labeled data for each class.  ...  To address this new task, we propose Prototype Network-based cross-attention contrastive learning (ProtoCACL) to capture the rich mutual interactions between the support set and query set.  ...  Compared with the main architecture of ProtoNets as shown in Figure 2 two novel parts, cross-attention and contrastive learning, as shown in Figure 2(b) . Prototypical Networks.  ... 
arXiv:2110.08254v1 fatcat:26utohj5ejenpospvnxtvcznnq

Multi-Level Matching and Aggregation Network for Few-Shot Relation Classification

Zhi-Xiu Ye, Zhen-Hua Ling
2019 Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics  
This paper presents a multi-level matching and aggregation network (MLMAN) for few-shot relation classification.  ...  Previous studies on this topic adopt prototypical networks, which calculate the embedding vector of a query instance and the prototype vector of each support set independently.  ...  ., 2019) added hybrid attention mechanism to prototypical networks, mainly focusing on improving the performance on few-shot relation classification with N > 1.  ... 
doi:10.18653/v1/p19-1277 dblp:conf/acl/YeL19 fatcat:33hop4rna5hnha7f76jlouqx3i

ReMP: Rectified Metric Propagation for Few-Shot Learning [article]

Yang Zhao, Chunyuan Li, Ping Yu, Changyou Chen
2020 arXiv   pre-print
Few-shot learning features the capability of generalizing from a few examples.  ...  to the success of metric-based few-shot learning.  ...  Performance illustration of popular few-shot learning models [5, 27, 28, 15, 24, 10] in 1-shot 5-way (Left) miniIma-geNet [31] and (Right) tieredImagenet classification.  ... 
arXiv:2012.00904v1 fatcat:rc2ld5mlgbbvxjaaiy6pr2rzba

Momentum Contrastive Learning for Few-Shot COVID-19 Diagnosis from Chest CT Images [article]

Xiaocong Chen and Lina Yao and Tao Zhou and Jinming Dong and Yu Zhang
2020 arXiv   pre-print
Specifically, we use contrastive learning to train an encoder which can capture expressive feature representations on large and publicly available lung datasets and adopt the prototypical network for classification  ...  To this end, we propose a new deep learning algorithm for the automated diagnosis of COVID-19, which only requires a few samples for training.  ...  Prototypical Network for Few-Shot Classification Another step in our workflow is classification.  ... 
arXiv:2006.13276v1 fatcat:up4hyb2qwre6xmbytxnfjo3l3u

CrossPoint: Self-Supervised Cross-Modal Contrastive Learning for 3D Point Cloud Understanding [article]

Mohamed Afham, Isuru Dissanayake, Dinithi Dissanayake, Amaya Dharmasiri, Kanchana Thilakarathna, Ranga Rodrigo
2022 arXiv   pre-print
Encouraged by this insight, we propose CrossPoint, a simple cross-modal contrastive learning approach to learn transferable 3D point cloud representations.  ...  Experimental results show that our approach outperforms the previous unsupervised learning methods on a diverse range of downstream tasks including 3D object classification and segmentation.  ...  (ii) Few-shot object classification. Few-shot learning (FSL) aims to train a model that generalizes with limited data.  ... 
arXiv:2203.00680v3 fatcat:xd33tvufxzb6ppwlo4rfy2em2m

Few-Shot Segmentation via Cycle-Consistent Transformer [article]

Gengwei Zhang, Guoliang Kang, Yi Yang, Yunchao Wei
2022 arXiv   pre-print
Few-shot segmentation aims to train a segmentation model that can fast adapt to novel classes with few exemplars.  ...  Directly performing cross-attention may aggregate these features from support to query and bias the query features.  ...  In a nutshell, our contributions are summarized as follows: ( 2 Related Work Few-Shot Segmentation Few-shot segmentation [26] is established to perform segmentation with very few exemplars.  ... 
arXiv:2106.02320v4 fatcat:bplork2sh5g3nctxlylpj3lvai

Learning Non-target Knowledge for Few-shot Semantic Segmentation [article]

Yuanwei Liu, Nian Liu, Qinglong Cao, Xiwen Yao, Junwei Han, Ling Shao
2022 arXiv   pre-print
First, a BG Mining Module (BGMM) is proposed to extract the BG region via learning a general BG prototype.  ...  Furthermore, we propose a prototypical contrastive learning algorithm to improve the model ability of distinguishing the target object from DOs.  ...  Furthermore, considering that it's hard to learn a good prototype feature embedding space to differentiate DOs from the target object under the few-shot setting, we propose the prototypical contrastive  ... 
arXiv:2205.04903v1 fatcat:sdcoywx7snex3n2ujmcocv34nq

Mining Latent Classes for Few-shot Segmentation [article]

Lihe Yang, Wei Zhuo, Lei Qi, Yinghuan Shi, Yang Gao
2021 arXiv   pre-print
Few-shot segmentation (FSS) aims to segment unseen classes given only a few annotated samples.  ...  both background and foreground categories to enforce more stable prototypes.  ...  Recently, contrastive learning based methods [14, 5] even perform on-par with the supervised counterparts in classification.  ... 
arXiv:2103.15402v3 fatcat:zksnaw3pqnaj5boieyqpdf6vsm

A Comparative Review of Recent Few-Shot Object Detection Algorithms [article]

Leng Jiaxu, Chen Taiyue, Gao Xinbo, Yu Yongtao, Wang Ye, Gao Feng, Wang Yue
2021 arXiv   pre-print
Few-shot object detection, learning to adapt to the novel classes with a few labeled data, is an imperative and long-lasting problem due to the inherent long-tail distribution of real-world data and the  ...  Finally, we conclude with the current status of few-shot object detection, along with potential research directions for this field.  ...  In recent years, few-shot learning has achieved several crucial breakthroughs, especially in few-shot classification (FSC) [14] - [27] .  ... 
arXiv:2111.00201v1 fatcat:preckuym7zarndm4yjc7n4k2oi

Dual Contrastive Learning for General Face Forgery Detection

Ke Sun, Taiping Yao, Shen Chen, Shouhong Ding, Jilin Li, Rongrong Ji
2022 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
Concretely, combined with the hard sample selection strategy, Inter-Instance Contrastive Learning (Inter-ICL) is first proposed to promote task-related discriminative features learning by especially constructing  ...  With various facial manipulation techniques arising, face forgery detection has drawn growing attention due to security concerns.  ...  Compared with other cross-entropy loss based methods, our two contrastive losses can both diverse the intra-category invariance and enhance the inconsistency of forgery face, thus the generalization can  ... 
doi:10.1609/aaai.v36i2.20130 fatcat:tldn6l3crbh6xdjzyqmpxbesmy

DeFRCN: Decoupled Faster R-CNN for Few-Shot Object Detection [article]

Limeng Qiao, Yuxuan Zhao, Zhiyuan Li, Xi Qiu, Jianan Wu, Chi Zhang
2021 arXiv   pre-print
classification model with taking the proposals from detector as input and boosting the original classification scores with additional pairwise scores for calibration.  ...  Few-shot object detection, which aims at detecting novel objects rapidly from extremely few annotated examples of previously unseen classes, has attracted significant research interest in the community  ...  As a recognition case of few-shot learning, few-shot classification has been widely investigated until now.  ... 
arXiv:2108.09017v1 fatcat:w3ntxyh3wbdq7bwrtovgy5x47u

Few-Shot Object Detection: A Survey

Simone Antonelli, Danilo Avola, Luigi Cinque, Donato Crisostomi, Gian Luca Foresti, Fabio Galasso, Marco Raoul Marini, Alessio Mecca, Daniele Pannone
2022 ACM Computing Surveys  
This survey, which to the best of our knowledge is the first tackling this problem, is focused on Few-Shot Object Detection, which has received far less attention compared to Few-Shot Classification due  ...  Few-Shot Learning aims at designing models which can effectively operate in a scarce data regime, yielding learning strategies that only need few supervised examples to be trained.  ...  uery Adaptive Few-Shot Object Detection with Heterogeneous Graph Convolutional Networks (QA-FewDet) [19] .  ... 
doi:10.1145/3519022 fatcat:trynjdbicvhezghv4k33n6xhfa

Prior Guided Feature Enrichment Network for Few-Shot Segmentation [article]

Zhuotao Tian, Hengshuang Zhao, Michelle Shu, Zhicheng Yang, Ruiyu Li, Jiaya Jia
2020 arXiv   pre-print
Few-shot segmentation is thus proposed to tackle this problem by learning a model that quickly adapts to new classes with a few labeled support samples.  ...  To alleviate these issues, we propose the Prior Guided Feature Enrichment Network (PFENet).  ...  Few-shot Learning Few-shot learning aims at image classification when only a few training examples are available.  ... 
arXiv:2008.01449v1 fatcat:5pary3mmzvbcjbfy3gm5gojqje

End-to-end One-shot Human Parsing [article]

Haoyu He, Bohan Zhuang, Jing Zhang, Jianfei Cai, Dacheng Tao
2022 arXiv   pre-print
Then, we propose learning momentum-updated prototypes by gradually smoothing the training time static prototypes, which helps stabilize the training and learn robust features.  ...  ., small sizes, testing bias, and similar parts, we devise an End-to-end One-shot human Parsing Network (EOP-Net).  ...  Next, we get the momentumupdated prototypes and perform prototype learning on the query features with a dual metric learning (DML) scheme, where we gradually shift network's focus from an Attention Guidance  ... 
arXiv:2105.01241v2 fatcat:o4cfib6tgfcmhaf5uyok3xdym4

Semantic-aligned Fusion Transformer for One-shot Object Detection [article]

Yizhou Zhao, Xun Guo, Yan Lu
2022 arXiv   pre-print
Specifically, we equip SaFT with a vertical fusion module (VFM) for cross-scale semantic enhancement and a horizontal fusion module (HFM) for cross-sample feature fusion.  ...  One-shot object detection aims at detecting novel objects according to merely one given instance.  ...  Being one of the underlying problems, few-shot learning received more and more interest from language [1, 17, 47, 59] to vision [15, 24, 37, 48, 50, 53, 55] related tasks.  ... 
arXiv:2203.09093v2 fatcat:tlsk6hk5vzbrjc4pp7pdnydo4e
« Previous Showing results 1 — 15 out of 2,477 results