Filters








81,599 Hits in 8.5 sec

Visual sentiment analysis via deep multiple clustered instance learning

Wenjing Gao, Wenjun Zhang, Haiyan Gao, Yonghua Zhu
2020 Journal of Intelligent & Fuzzy Systems  
We propose a deep multiple clustered instance learning formulation, under which a deep multiple clustered instance learning network (DMCILN) is constructed for visual sentiment analysis.  ...  Based on the observation that visual sentiment is conveyed through many visual elements in images, we put forward to tackle visual sentiment analysis under multiple instance learning (MIL) formulation.  ...  et al. / Visual sentiment analysis via deep multiple clustered instance learning Fig. 2 . 2 Distinct problem formulations and learning goals among classical supervised learning, MIL, and MCIL.  ... 
doi:10.3233/jifs-200675 fatcat:ix3hqllr7vdb5b76nvtdt6lyhe

Detecting cutaneous basal cell carcinomas in ultra-high resolution and weakly labelled histopathological images [article]

Susanne Kimeswenger, Elisabeth Rumetshofer, Markus Hofmarcher, Philipp Tschandl, Harald Kittler, Sepp Hochreiter, Wolfram Hötzenecker, Günter Klambauer
2019 arXiv   pre-print
Increasingly, machine learning methods are applied due to their superior performance.  ...  and the weak labels of whole slide images.  ...  We thank the NVIDIA Corporation, Audi.JKU Deep Learning Center, Audi Electronic Venture GmbH, Janssen Pharmaceutica (MadeSMART), UCB S.A., FFG grant 871302, LIT grant DeepToxGen and AI-SNN, and FWF grant  ... 
arXiv:1911.06616v3 fatcat:ua45qarto5hsposvzt5cwqtwrm

ML-LocNet: Improving Object Localization with Multi-view Learning Network [chapter]

Xiaopeng Zhang, Yang Yang, Jiashi Feng
2018 Lecture Notes in Computer Science  
The multi-view learning would benefit localization due to the complementary relationships among the learned features from different views and the consensus property among the mined instances from each  ...  This paper addresses Weakly Supervised Object Localization (WSOL) with only image-level supervision.  ...  Among these optimization strategies, a typical solution is to alternate between model re-training and object re-localization, which shares a similar spirit with Multiple Instance Learning (MIL) [5] ,  ... 
doi:10.1007/978-3-030-01219-9_15 fatcat:swmcsa5eije77kd6xb625bqfvu

Graphonomy: Universal Image Parsing via Graph Reasoning and Transfer [article]

Liang Lin and Yiming Gao and Ke Gong and Meng Wang and Xiaodan Liang
2021 arXiv   pre-print
In particular, Graphonomy learns the global and structured semantic coherency in multiple domains via semantic-aware graph reasoning and transfer, enforcing the mutual benefits of the parsing across domains  ...  ) without extensive re-training.  ...  domains via graph transfer learning to achieve multiple levels of human parsing tasks.  ... 
arXiv:2101.10620v1 fatcat:hnbuqiugsfhvbc7phn5htmsvcy

Keywords to visual categories: Multiple-instance learning forweakly supervised object categorization

Sudheendra Vijayanarasimhan, Kristen Grauman
2008 2008 IEEE Conference on Computer Vision and Pattern Recognition  
We show that multiple-instance learning enables the recovery of robust category models from images returned by keyword-based search engines.  ...  Conventional supervised methods for image categorization rely on manually annotated (labeled) examples to learn good object models, which means their generality and scalability depends heavily on the amount  ...  (a) Given a category name, our method automatically collects noisy "positive bags" of instances via keyword-based image search on multiple search engines in multiple languages.  ... 
doi:10.1109/cvpr.2008.4587632 dblp:conf/cvpr/VijayanarasimhanG08 fatcat:mfoghhdbrnbi3ofs7ensu6ep2m

Deep Learning for Instance Retrieval: A Survey [article]

Wei Chen, Yu Liu, Weiping Wang, Erwin Bakker, Theodoros Georgiou, Paul Fieguth, Li Liu, Michael S. Lew
2022 arXiv   pre-print
In this survey we review recent instance retrieval works that are developed based on deep learning algorithms and techniques, with the survey organized by deep network architecture types, deep features  ...  Artificial intelligence has made progress in CBIR and has significantly facilitated the process of instance search.  ...  ACKNOWLEDGMENT The authors would like to thank the pioneer researchers in instance retrieval and other related fields.  ... 
arXiv:2101.11282v3 fatcat:qvodunmw4bdltcneadyt7d7h5m

MIST: Multiple Instance Spatial Transformer Network [article]

Baptiste Angles and Yuhe Jin and Simon Kornblith and Andrea Tagliasacchi and Kwang Moo Yi
2020 arXiv   pre-print
We propose a deep network that can be trained to tackle image reconstruction and classification problems that involve detection of multiple object instances, without any supervision regarding their whereabouts  ...  Our method is able to learn to detect recurrent structures in the training dataset by learning to reconstruct images.  ...  Wan et al. (2018) learn how to detect multiple instances of a single object via region proposals and ROI pooling, while Tang et al. (2018) propose to use a hierarchical setup to refine their estimates  ... 
arXiv:1811.10725v5 fatcat:qm47oychvffn3jd6lzejtaz5xu

Region-DH: Region-based Deep Hashing for Multi-Instance Aware Image Retrieval

Franck Romuald Fotso Mtope, Bo Wei
2020 2020 International Joint Conference on Neural Networks (IJCNN)  
This paper introduces an instance-aware hashing approach Region-DH for large-scale multi-label image retrieval.  ...  Region-DH can flexibly be used with existing deep neural networks or more complex object detectors for image hashing.  ...  It also determines hashing functions via both instance-level hashing (H1) and image-level hashing (H2).  ... 
doi:10.1109/ijcnn48605.2020.9207485 dblp:conf/ijcnn/MtopeW20 fatcat:zvyrgns65vfabgyui6sy2rxiwa

Pornographic Image Recognition via Weighted Multiple Instance Learning [article]

Jin Xin, Wang Yuhui, Tan Xiaoyang
2019 arXiv   pre-print
In this paper, we model each image as a bag of regions, and follow a multiple instance learning (MIL) approach to train a generic region-based recognition model.  ...  ., breast and private part) in an image often lie in local regions of small size.  ...  Multiple Instance Learning Research on Multiple Instance Learning (MIL) studies the problem where a real-world object is associated with a single class label but described by a number of instances.  ... 
arXiv:1902.03771v1 fatcat:x6almdzggrgepl3lsulepe6mcu

A Multiple Component Matching Framework for Person Re-identification [chapter]

Riccardo Satta, Giorgio Fumera, Fabio Roli, Marco Cristani, Vittorio Murino
2011 Lecture Notes in Computer Science  
Building on these similarities, we propose a Multiple Component Matching (MCM) framework for the person reidentification problem, which is inspired by Multiple Component Learning, a framework recently  ...  Despite this, previous works share similar representations of human body based on part decomposition and the implicit concept of multiple instances.  ...  An ensemble of MIL classifiers, one for each region, is trained via a boosting algorithm, so that the most discriminative regions get a higher weight.  ... 
doi:10.1007/978-3-642-24088-1_15 fatcat:qopggoaj6vbnvcfj4b6pfwpjzi

A Multiple Component Matching Framework for Person Re-Identification [article]

Riccardo Satta, Giorgio Fumera, Fabio Roli, Marco Cristani and Vittorio Murino
2011 arXiv   pre-print
Building on these similarities, we propose a Multiple Component Matching (MCM) framework for the person re-identification problem, which is inspired by Multiple Component Learning, a framework recently  ...  Despite this, previous works share similar representations of human body based on part decomposition and the implicit concept of multiple instances.  ...  An ensemble of MIL classifiers, one for each region, is trained via a boosting algorithm, so that the most discriminative regions get a higher weight.  ... 
arXiv:1105.2491v2 fatcat:pfbthrze65b47a45y7fe5urexq

Zigzag Learning for Weakly Supervised Object Detection [article]

Xiaopeng Zhang, Jiashi Feng, Hongkai Xiong, Qi Tian
2018 arXiv   pre-print
instances properly.  ...  These two modules formulate a zigzag learning process, where progressive learning endeavors to discover reliable object instances, and masking regularization increases the difficulty of finding object  ...  Among them, a typical solution is alternating between model re-training and object re-localization, which shares a similar spirit with Multiple Instance Learning (MIL) [9] , [10] , [11] .  ... 
arXiv:1804.09466v1 fatcat:22l3yozl6rfv5mazkvute2ohjq

Zigzag Learning for Weakly Supervised Object Detection

Xiaopeng Zhang, Jiashi Feng, Hongkai Xiong, Qi Tian
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
instances properly.  ...  These two modules formulate a zigzag learning process, where progressive learning endeavors to discover reliable object instances, and masking regularization increases the difficulty of finding object  ...  Among them, a typical solution is alternating between model re-training and object re-localization, which shares a similar spirit with Multiple Instance Learning (MIL) [9] , [10] , [11] .  ... 
doi:10.1109/cvpr.2018.00448 dblp:conf/cvpr/0008FX018 fatcat:y6jn5b5o65ak5fh2yubn57ey5q

Rescue Tail Queries

Xiaopeng Yang, Tao Mei, Yongdong Zhang
2014 Proceedings of the ACM International Conference on Multimedia - MM '14  
The approach can not only effectively expand training data by learning relevant information from the constructed click-wise-based image pairs, but also fully explore the effects of multiple visual modalities  ...  We therefore propose a novel query-dependent learning to re-rank approach for tail queries, called "click-wise multimodal fusion."  ...  This further shows the superiority of relevant information mined from click-wise-based image pairs used in our approach.  ... 
doi:10.1145/2647868.2654900 dblp:conf/mm/YangMZ14 fatcat:plpat2pfrjgqtbqwiad4whqowi

Spatial Cross-Attention Improves Self-Supervised Visual Representation Learning [article]

Mehdi Seyfi, Amin Banitalebi-Dehkordi, Yong Zhang
2022 arXiv   pre-print
We can later remove the add-on module for inference without any modification of the learned weights.  ...  The main idea behind these methods is that different views of a same image represent the same semantics.  ...  We also set the loss weights to be w 0 = 1, w 1 = .05, w 2 = .1. Unsupervised Training : We train the networks using LARC optimizer with momentum of .9 and weight decay of 1e − 6.  ... 
arXiv:2206.05028v1 fatcat:zfzelprk4fambjhrmxnmjrk7q4
« Previous Showing results 1 — 15 out of 81,599 results