The DOI matches its Crossref record; 56 of the 57 references it cites were found in a registry or an index.
Hide-and-Seek: Forcing a Network to be Meticulous for Weakly-Supervised Object and Action Localization
Open accessAbstract
We propose `Hide-and-Seek', a weakly-supervised framework that aims to improve object localization in images and action localization in videos. Most existing weakly-supervised methods localize only the most discriminative parts of an object rather than all relevant parts, which leads to suboptimal performance. Our key idea is to hide patches in a training image randomly, forcing the network to seek other relevant parts when the most discriminative part is hidden. Our approach only needs to modify the input image and can work with any network designed for object localization. During testing, we do not need to hide any patches. Our Hide-and-Seek approach obtains superior performance compared to previous methods for weakly-supervised object localization on the ILSVRC dataset. We also demonstrate that our framework can be easily extended to weakly-supervised action localization.
Overview
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“However, various types of industrial image defects do not meet all the above assumptions [42,43].”
From An Improved VGG19 Transfer Learning Strip Steel Surface Defect Recognition Deep Neural Network Based on Few Samples and Imbalanced Datasets (2021) · cites this paper as [43]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
ImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155
Going deeper with convolutions
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298594 · Cited by 48,562
ImageNet Large Scale Visual Recognition Challenge
Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. (2015) · International Journal of Computer Vision · doi:10.1007/s11263-015-0816-y · Cited by 39,380
Fully convolutional networks for semantic segmentation
Long, J., Shelhamer, E., & Darrell, T. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298965 · Cited by 38,834
(2014) · Cited by 33,419
Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation
Girshick, R., Donahue, J., Darrell, T., & Malik, J. (2014) · 2014 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2014.81 · Cited by 32,684
(2016) · Cited by 31,590
Girshick, R. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.169 · Cited by 28,105
Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
Ioffe, S., & Szegedy, C. (2015) · arXiv · doi:10.48550/arxiv.1502.03167 · Cited by 27,818
Visualizing and understanding convolutional networks
Zeiler, M. D., & Fergus, R. (2014) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-10590-1_53 · Cited by 15,839
Flagged first, then by citations
1 of the 57 references couldn’t be matched to a known paper.
Research works printed but not linked: 5
- jiang, 2014, THUMOS Challenge Action Recognition with A Large Number of Classes
- jhuang, 2013, Towards undestanding action recognition, ICCV
- 10.1109/CVPR.2016.311 · DOI · Paper page
- cinbis, 2015, Weakly supervised object localization with multi-fold multiple instance learning
- wan, 2013, Regularization of neural network using dropconnect, ICML
Cited by
Sorted by citations
Items 101–200 of 720
Deep Learning-Based Action Detection in Untrimmed Videos: A Survey
Vahdani, E., & Tian, Y. (2022) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2022.3193611 · Cited by 100
“WS Hide-Seek [76] Completeness with RandomMasking 19.”
BackgroundMethodInfluentialCorrelation Verification for Image Retrieval
Lee, S., Seong, H., Lee, S., & Kim, E. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.00530 · Cited by 90
“Similarly, to prove the effectiveness of the Hide-and-Seek [13] augmentation, we conduct experiments by varying the Hide-and-Seek probability phas.”
InfluentialWeakly Supervised Temporal Adjacent Network for Language Grounding
Wang, Y., Deng, J., Zhou, W., & Li, H. (2021) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2021.3096087 · Cited by 88
Min-Entropy Latent Model for Weakly Supervised Object Detection
Wan, F., Wei, P., Han, Z., Jiao, J., & Ye, Q. (2019) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2019.2898858 · Cited by 95
“the convolutional filters behave as detectors to activate regions of interest on the deep feature maps [39], [40], [41].”
BackgroundLarge-Scale Unsupervised Semantic Segmentation
Gao, S., Li, Z.-Y., Yang, M.-H., Cheng, M.-M., Han, J., & Torr, P. (2023) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2022.3218275 · Cited by 88
UniTAB: Unifying Text and Box Outputs for Grounded Vision-Language Modeling
Yang, Z., Gan, Z., Wang, J., Hu, X., Ahmed, F., Liu, Z., Lu, Y., & Wang, L. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-20059-5_30 · Cited by 105
“The “GT-known” [57,75,76,15] setting aims to localize a given ground-truth class.”
BackgroundVectorized Evidential Learning for Weakly-Supervised Temporal Action Localization
Gao, J., Chen, M., & Xu, C. (2023) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2023.3311447 · Cited by 81
“Most existing WS-TAL approaches adopt the localization-by-classification framework [5], [7], [18], [39], in which the commonly used multiple-instance learning (MIL) strategy [40], [41], [42], attention mechanism [19], [39], [43], [44], or erasing-based strategy [45], [46] are employed.”
MethodLearning Action Completeness from Points for Weakly-supervised Temporal Action Localization
Lee, P., & Byun, H. (2021) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.01339 · Cited by 79
“Moreover, some approaches aim to detect complete action instances by aggregating multiple predictions [25], erasing the most discriminative part [44, 63], or directly regressing the action intervals [28, 42].”
BackgroundRossetti, S., Zappia, D., Sanzari, M., Schaerf, M., & Pirri, F. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-20056-4_26 · Cited by 80
MCTformer+: Multi-Class Token Transformer for Weakly Supervised Semantic Segmentation
Xu, L., Bennamoun, M., Boussaïd, F., Laga, H., Ouyang, W., & Xu, D. (2024) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2024.3404422 · Cited by 93
“This has been implemented through various modifications to either the input data [17], [18], [19], [20], [21] or feature maps [22], [23], [24], including techniques like dropping out certain image information or introducing perturbations.”
MethodDual-Evidential Learning for Weakly-supervised Temporal Action Localization
Chen, M., Gao, J., Yang, S., & Xu, C. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-19772-7_12 · Cited by 80
“To date in the literature, most existing approaches can be divided into three categories, namely attention-based methods [45, 54, 13, 42, 38, 32], MIL-based methods [35, 22, 33, 37, 40], and erasing-based methods [48, 59, 61].”
BackgroundFigueroa, K. C., Song, B., Sunny, S. P., Li, S., Gurushanth, K., Mendonca, P., Mukhia, N., Patrick, S., Gurudath, S., Raghavan, S., et al. (2022) · Journal of Biomedical Optics · doi:10.1117/1.jbo.27.1.015001 · Cited by 84
Out-of-Candidate Rectification for Weakly Supervised Semantic Segmentation
Cheng, Z., Qiao, P., Li, K., Li, S., Wei, P., Ji, X., Li, Y., Liu, C., & Chen, J. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.02267 · Cited by 78
“A few methods design heuristic strategies, like "Hide & Seek" [53] and Erasing [63], adopted on images [37, 74, 77] or feature maps [12, 27, 33] to force the network to exploit novel regions rather than only discriminative regions.”
MethodTowards Interpretable Face Recognition
Yin, B., Tran, L., Li, H., Shen, X., & Liu, X. (2019) · 2019 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00944 · Cited by 90
SQUID: Deep Feature In-Painting for Unsupervised Anomaly Detection
Xiang, T., Zhang, Y., Lu, Y., Yuille, A., Zhang, C., Cai, W., & Zhou, Z. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.02288 · Cited by 74
“It is understandable because our SQUID is an unsupervised method, requiring zero manual annotation for normal/abnormal images, unlike [60, 68, 72, 75, 80].”
MethodRegularization Strategy for Point Cloud via Rigidly Mixed Sample
Lee, D., Lee, J., Lee, J., Lee, H., Lee, M., Woo, S., & Lee, S. (2021) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr46437.2021.01564 · Cited by 79
“Some works have enabled the model to learn spatially distributed representation by removing the part of the data on pixel [36, 23, 8] or feature map [9] basis.”
BackgroundMethodLesion-attention pyramid network for diabetic retinopathy grading
Li, X., Jiang, Y., Zhang, J., Li, M., Luo, H., & Yin, S. (2022) · Artificial Intelligence in Medicine · doi:10.1016/j.artmed.2022.102259 · Cited by 73
LiSiam: Localization Invariance Siamese Network for Deepfake Detection
Wang, J., Sun, Y., & Tang, J. (2022) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2022.3186803 · Cited by 73
Hide-and-Seek: A Data Augmentation Technique for Weakly-Supervised Localization and BeyondPreprint
Singh, K. K., Yu, H., Sarmasi, A., Pradeep, G., & Lee, Y. J. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1811.02545 · Cited by 96
“This paper expands upon our previous conference paper [24].”
Self-citationLearning Temporal Co-Attention Models for Unsupervised Video Action Localization
Gong, G., Wang, X., Mu, Y., & Tian, Q. C. (2020) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr42600.2020.00984 · Cited by 71
“UntrimmedNet [46], Hide-and-Seek [41], W-TALC [29], Sparse Temporal Pooling Network [27], AutoLoc [38], and CleanNet[14], etc.”
MethodTemporal Action Localization in the Deep Learning Era: A Survey
Wang, B., Zhao, Y., Yang, L., Long, T., & Li, X. (2023) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2023.3330794 · Cited by 79
“Firstly, the “hide-and-seek” approach [57], along with its extension [146], intentionally conceals discriminative action parts during training.”
BackgroundMethodInfluentialXie, J., Luo, C., Zhu, X., Jin, Z., Lu, W., & Shen, L. (2021) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.00020 · Cited by 71
“The random erase of image patches [27] helps to locate less discriminative regions of objects.”
InfluentialZhang, X., Shi, H., Li, C., Zheng, K., Zhu, X., & Duan, L. (2019) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v33i01.33019227 · Cited by 69
Belharbi, S., Rony, J., Dolz, J., Ayed, I. B., McCaffrey, L. M., & Granger, É. (2021) · IEEE Transactions on Medical Imaging · doi:10.1109/tmi.2021.3123461 · Cited by 73
Relational Prototypical Network for Weakly Supervised Temporal Action Localization
Huang, L., Huang, Y., Ouyang, W., & Wang, L. (2020) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v34i07.6760 · Cited by 68
D2-Net: Weakly-Supervised Action Localization via Discriminative Embeddings and Denoised Activations
Narayan, S., Cholakkal, H., Hayat, M., Khan, F. S., Yang, M., & Shao, L. (2021) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.01335 · Cited by 70
“Several weak supervision strategies have been explored in the context of action localization, including category labels [33, 23, 34, 25, 31, 41], sparse temporal points [19], order of actions [26, 2], instance count [22, 39] and singleframe annotations [17].”
BackgroundMethodForeground Activation Maps for Weakly Supervised Object Localization
Meng, M., Zhang, T., Tian, Q., Zhang, Y., & Wu, F. (2021) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.00337 · Cited by 70
A Survey on Temporal Action Localization
Xia, H., & Zhan, Y. (2020) · IEEE Access · doi:10.1109/access.2020.2986861 · Cited by 75
“To solve the problem of fragmentation of video frames that the classifier cares about and the completeness of the action, Hide-and-seek [30] randomly hides some frames to force residual attention to learn the relatively low discrimination video frames in each training.”
BackgroundMethodZhang, S., Qu, N., Zheng, T., & Hu, C. (2022) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2022.3158990 · Cited by 68
Leveraging the Present to Anticipate the Future in Videos
Miech, A., Laptev, I., Šivic, J., Wang, H., Torresani, L., & Tran, D. (2019) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw.2019.00351 · Cited by 68
“Such advances have manifested in disparate video understanding tasks, including action recognition [5, 8, 11, 38, 41], temporal action localization [37, 39, 47, 53], video search [13], video summarization [32] and video categorization [29].”
BackgroundFan, D.-P., Zhang, J., Xu, G., Cheng, M., & Shao, L. (2022) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2022.3166451 · Cited by 74
“Three different strategies have been widely studied, including label smoothing [294], image augmentation [295], [296], and self-supervised learning [297].”
BackgroundMethodRadiologist-Level COVID-19 Detection Using CT Scans with Detail-Oriented Capsule NetworksPreprint
Mobiny, A., Cicalese, P. A., Zare, S., Yuan, P., Abavisani, M., Wu, C. C., Ahuja, J., de Groot, P. M., & Nguyen, H. V. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2004.07407 · Cited by 70
“It is inspired by the Hide and Seek [33] mechanism in which image patches are randomly dropped to encourage the network to look for other relevant parts.”
MethodCircumventing Outliers of AutoAugment with Knowledge Distillation
Wei, L., Xiao, A., Xie, L., Zhang, X., Chen, X., & Tian, Q. C. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-58580-8_36 · Cited by 68
Image data augmentation techniques based on deep learning: A survey
Zeng, W. (2024) · Mathematical Biosciences & Engineering · doi:10.3934/mbe.2024272 · Cited by 93
Adversarial Seeded Sequence Growing for Weakly-Supervised Temporal Action Localization
Zhang, C., Xu, Y., Cheng, Z., Niu, Y. F., Pu, S., Wu, F. Q., & Zou, F. (2019) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3343031.3351044 · Cited by 64
“ong 59.8 57.1 53.2 48.5 42.8 – – Alwasssel et al.[1] strong 49.6 44.3 38.1 28.4 19.8 – – BSN [14] strong – – 53.5 45.0 36.9 33.8 20.8 UntrimmedNet [24] weak 44.4 37.7 28.2 21.1 13.7 – – Hide-and-Seek [23] weak 36.4 27.8 19.5 12.7 6.8 – – Zhong et al. [31] weak 45.8 39.0 31.1 22.5 15.9 – – AutoLoc [20] weak – – 35.8 29…”
BackgroundInfluentialBidirectional Attention-Recognition Model for Fine-Grained Object Classification
Liu, C., Xie, H., Zha, Z.-J., Yu, L., Chen, Z., & Zhang, Y. (2019) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2019.2954747 · Cited by 63
“[58] put forward Hide-and-Seek to hide image patches randomly then force the CNN to seek other parts.”
BackgroundStrengthen Learning Tolerance for Weakly Supervised Object Localization
Guo, G., Han, J., Wan, F., & Zhang, D. (2021) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr46437.2021.00732 · Cited by 66
“[53] 64.4 42.2 54.7 - GoogLeNet-CAM [53] 65.0 43.6 57.0 - GoogLeNet-HaS-32 [28] 70.7 45.2 - 60.3 GoogLeNet-ACoL [48] 71.0 46.7 57.4 - InceptionV3-CAM [53] 73.3 46.3 58.2 62.7 InceptionV3-SPG [49] - 48.6 60.0 - InceptionV3-DANet [40] 72.5 47.5 58.3 - InceptionV3-ADL [6] 72.8 48.7 - -…”
BackgroundMethodResultInfluentialUnderstanding Deep Networks via Extremal Perturbations and Smooth Masks
Fong, R., Patrick, M., & Vedaldi, A. (2019) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00304 · Cited by 89
“Other works have leveraged perturbations at the input [23, 29] and intermediate layers [28] to perform weakly and fully supervised localization.”
BackgroundWeakly Supervised Action Selection Learning in Video
Ma, J., Gorti, S. K., Volkovs, M., & Yu, G. (2021) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr46437.2021.00750 · Cited by 63
Rony, J., Belharbi, S., Dolz, J., Ayed, I. B., McCaffrey, L. M., & Granger, É. (2023) · The Journal of Machine Learning for Biomedical Imaging · doi:10.59275/j.melba.2023-5g54 · Cited by 83
“, 2016), HaS (Singh and Lee, 2017), WILDCAT (Durand et al.”
BackgroundMethodMaskCOV: A random mask covariance network for ultra-fine-grained visual categorization
Yu, X., Zhao, Y. S., Gao, Y., & Xiong, S. (2021) · Pattern Recognition · doi:10.1016/j.patcog.2021.108067 · Cited by 65
Weakly-Supervised Learning for Tool Localization in Laparoscopic Videos
Vardazaryan, A., Mutter, D., Marescaux, J., & Padoy, N. (2018) · Lecture notes in computer science · doi:10.1007/978-3-030-01364-6_19 · Cited by 60
Zhang, X., Li, C., Shi, H., Zhu, X., Li, P., & Dong, J. (2020) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2019.2962815 · Cited by 59
“Weak Singh and Lee, 2017 [59] 36.”
Guided Attention Inference Network
Li, K., Wu, Z., Peng, K., Ernst, J., & Fu, Y. (2019) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2019.2921543 · Cited by 59
“Although the attention maps are supervised by only classification loss without pixel-level labels, the attention maps generated by the trained network usually cover only the small and discriminative areas of the object of interest [22], [43], [59].”
MethodLCTR: On Awakening the Local Continuity of Transformer for Weakly Supervised Object Localization
Chen, Z., Wang, C., Wang, Y., Jiang, G., Shen, Y., Tai, Y. Y., Wang, C., Zhang, W., & Cao, L. (2022) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v36i1.19918 · Cited by 58
Exploring Denoised Cross-video Contrast for Weakly-supervised Temporal Action Localization
Li, J., Yang, T., Ji, W., Wang, J., & Cheng, L. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.01929 · Cited by 58
“To facilitate the detection of complete actions, [33, 46, 69] propose to remove the discriminative action parts or randomly hide”
Self-supervised object detection from audio-visual correspondence
Afouras, T., Asano, Y. M., Fagan, F., Vedaldi, A., & Metze, F. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.01032 · Cited by 56
“Other methods use technique such as CAM or analogous techniques [15, 19, 27, 80, 84, 110] as a form of weakly supervised saliency or localisation maps.”
MethodWeakly Supervised Temporal Action Localization Using Deep Metric Learning
Islam, A., & Radke, R. J. (2020) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv45572.2020.9093620 · Cited by 52
“Weak Hide-and-Seek [36] 36.”
Siamese Alignment Network for Weakly Supervised Video Moment Retrieval
Wang, Y., Liu, M., Wei, Y., Cheng, Z., Wang, Y., & Nie, L. (2022) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2022.3168424 · Cited by 54
Bridging the Gap between Classification and Localization for Weakly Supervised Object Localization
Kim, E., Kim, S., Lee, J., Kim, H., & Yoon, S. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.01386 · Cited by 54
“HaS [21] trains a classifier using Input Image”
BackgroundMethodBackground Activation Suppression for Weakly Supervised Object Localization
Wu, P., Zhai, W., & Cao, Y. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.01385 · Cited by 54
“HaS [16] randomly splits lhe original image into different patches, forcing the classification network to learn more features of objects.”
BackgroundMethodSmoothMix: a Simple Yet Effective Data Augmentation to Train Robust Classifiers
Lee, J. H., Zaheer, M. Z., Astrid, M., & Lee, S. (2020) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw50498.2020.00386 · Cited by 55
“Table 3 provides the image classification performance results of SmoothMixC and its comparison against state-of-the-art methodologies [10, 45, 22, 58, 51, 7, 25, 15, 56].”
BackgroundMethodInfluentialAction Graphs: Weakly-supervised Action Localization with Graph Convolution Networks
Rashid, M., Kjellström, H., & Lee, Y. J. (2020) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv45572.2020.9093404 · Cited by 51
“Others discourage the network from focusing only on the most discriminative time segments via random hiding [28], or their iterative removal during training [38].”
BackgroundSelf-citationSet-Constrained Viterbi for Set-Supervised Action Segmentation
Li, J., & Todorović, S. (2020) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr42600.2020.01083 · Cited by 50
Partial Class Activation Attention for Semantic Segmentation
Liu, S.-A., Xie, H., Xu, H. Y., Zhang, Y., & Tian, Q. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.01633 · Cited by 52
“Methods like [23, 28] propose region erasing to enlarge the activated area in each CAM.”
Silva-Rodríguez, J., Colomer, A., & Naranjo, V. (2021) · Computerized Medical Imaging and Graphics · doi:10.1016/j.compmedimag.2020.101846 · Cited by 52
“During the training stage, two techniques are carried out to regularize the model and avoid over-fitting: data augmentation and hide-and-seek [35].”
MethodPivoTAL: Prior-Driven Supervision for Weakly-Supervised Temporal Action Localization
Rizve, M. N., Mittal, G. K., Yu, Y., Hall, M., Sajeev, S., Shah, M. A., & Chen, M. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.02202 · Cited by 54
“The single-stage WTAL methods can be further divided into three main categories: MIL-based [20, 33, 35, 40], attention-based [11, 31, 36, 41, 45, 50], and erasingbased [19, 34, 55, 61] methods.”
MethodUSAGE: A Unified Seed Area Generation Paradigm for Weakly Supervised Semantic Segmentation
Peng, Z., Wang, G., Xie, L., Jiang, D., Shen, W., & Tian, Q. C. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00064 · Cited by 51
“various types of weak supervision, we focus on studying WSSS under image-level labels [12, 20, 17, 18, 10, 37], which is considered one of the most challenging scenarios and has attracted increasing attention.”
BackgroundJang, Y., Güneş, H., & Patras, I. (2019) · Computer Vision and Image Understanding · doi:10.1016/j.cviu.2019.01.006 · Cited by 54
“in the horizontal direction with a probability of 0.5. In the training session, we randomly select one of data augmentation mechanisms (shrinking, cropping, gamma correction and Hide-and-Seek (H-a-S) [47]) to create noisy data samples for each epoch. Both shrinking and cropping maintain the aspect ratio. Gamma correct…”
MethodInfluentialFeature mask network for person re-identification
Ding, G., Khan, S., Tang, Z., & Porikli, F. (2019) · Pattern Recognition Letters · doi:10.1016/j.patrec.2019.02.015 · Cited by 53
RefineLoc: Iterative Refinement for Weakly-Supervised Action Localization
Pardo, A., Alwassel, H., Heilbron, F. C., Thabet, A., & Ghanem, B. (2021) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv48630.2021.00336 · Cited by 52
Exploring Sub-Action Granularity for Weakly Supervised Temporal Action Localization
Wang, B., Zhang, X., & Zhao, Y. (2021) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2021.3089323 · Cited by 49
VideoMix: Rethinking Data Augmentation for Video ClassificationPreprint
Yun, S., Oh, S. J., Heo, B., Han, D., & Kim, J. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2012.03457 · Cited by 61
“9, and we report their mean value (mAP) as in [29, 22, 20].”
BackgroundResultInfluentialThe Blessings of Unlabeled Background in Untrimmed Videos
Liu, Y., Chen, J., Chen, Z., Deng, B., Huang, J., & Zhang, H. (2021) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr46437.2021.00611 · Cited by 49
“Many approaches have been proposed to tackle the issues by erasing salient features [36, 47], imposing diversity loss [16], and employing marginalized average aggregation strategy [41].”
BackgroundAdaptive Patch Contrast for Weakly Supervised Semantic Segmentation
Wu, W., Dai, T., Chen, Z., Huang, X., Xiao, J., Ma, F., & Ouyang, R. (2024) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2024.109626 · Cited by 54
Wang, C., Caragea, D., Narayana, N. K., Hein, N. T., Bheemanahalli, R., Somayanda, I. M., & Jagadish, S. V. K. (2022) · Plant Methods · doi:10.1186/s13007-022-00839-5 · Cited by 51
Detection of Foreign Objects Intrusion Into Transmission Lines Using Diverse Generation Model
Wu, Y., Zhao, S., Xing, Z., Zheng, W., Li, Y., & Li, Y. (2023) · IEEE Transactions on Power Delivery · doi:10.1109/tpwrd.2023.3279891 · Cited by 47
Meng, F., Luo, K., Li, H., Wu, Q., & Xu, X. (2019) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2019.2962073 · Cited by 47
“Designing efficient segmentation loss is essential to the one-stage-based method, and several types of segmentation losses have been studied, such as the classification loss by image-level labels [11], [12] and the segmentation loss by pseudo-annotations [15].”
MethodACM-Net: Action Context Modeling Network for Weakly-Supervised Temporal Action LocalizationPreprint
Qu, S., Chen, G., Li, Z., Zhang, L., Fan, L., & Knoll, A. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2104.02967 · Cited by 59
Fine-Grained Hashing With Double Filtering
Chen, Z.-D., Luo, X., Wang, Y., Guo, S., & Xu, X.-S. (2022) · IEEE Transactions on Image Processing · doi:10.1109/tip.2022.3145159 · Cited by 46
Weakly Supervised Object Localization via Transformer with Implicit Spatial Calibration
Bai, H., Zhang, R., Wang, J., & Wan, X. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-20077-9_36 · Cited by 46
“For instance, HaS [28] hides patches in training images randomly to force the network to seek other relevant parts; CutMix [34] adopts the same way to drop out patches but further augment the area of the patches with ground-truth labels to reduce information loss.”
BackgroundMethodCREAM: Weakly Supervised Object Localization via Class RE-Activation Mapping
Xu, J., Hou, J., Zhang, Y., Feng, R., Zhao, R.-W., Zhang, T., Lu, X., & Gao, S. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.00922 · Cited by 49
“We integrate CREAMBase into HAS and SPG, resulting in HAS CREAMBase and SPG CREAMBase, respectively.”
BackgroundMethodInfluentialYang, S., Guo, S., Zhao, J., & Shen, F. (2023) · Pattern Recognition · doi:10.1016/j.patcog.2023.110204 · Cited by 50
Treating Pseudo-labels Generation as Image Matting for Weakly Supervised Semantic Segmentation
Wang, C., Xu, R., Xu, S., Meng, W., & Zhang, X. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00076 · Cited by 45
“Erasure-based approaches [14, 20, 46, 44] use iterative erasure strategies that retain erased features to prevent the classification network from focusing only on discriminative object parts.”
MethodLearning to Exploit the Prior Network Knowledge for Weakly Supervised Semantic Segmentation
Redondo-Cabrera, C., Baptista-Rios, M., & López-Sastre, R. J. (2019) · IEEE Transactions on Image Processing · doi:10.1109/tip.2019.2901393 · Cited by 45
“In [27], a hide and seek module is also introduced to train the original CAM of [16] using random hidden image patches to increase the precision of the class activation maps.”
BackgroundMethodInfluentialAction Recognition via Pose-Based Graph Convolutional Networks with Intermediate Dense Supervision
Shi, L., Zhang, Y., Cheng, J., & Lu, H.-Q. (2019) · Pattern Recognition · doi:10.1016/j.patcog.2021.108170 · Cited by 34
ResizeMix: Mixing Data with Preserved Object Information and True LabelsPreprint
Qin, J., Fang, J., Zhang, Q., Liu, W., Wang, X., & Wang, X. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2012.11101 · Cited by 59
“Cutting- and Mixing- based Data Augmentation The goal of cutting augmentations is to make a network pay attention to the entire data like the dropout regularization [7, 18, 20, 46, 47].”
What Can Human Sketches Do for Object Detection?
Chowdhury, P. N., Bhunia, A. K., Sain, A., Koley, S., Xiang, T., & Song, Y.-Z. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.01448 · Cited by 45
“Approaches vary from simple image rotation and flipping to more advanced techniques of image erasing [11] like CutOut [18], Hide-and-Seek [81] and image mixing like MixUp [102] and CutMix [101].”
MethodSPARE: Self-supervised part erasing for ultra-fine-grained visual categorization
Yu, X., Zhao, Y., & Gao, Y.-S. (2022) · Pattern Recognition · doi:10.1016/j.patcog.2022.108691 · Cited by 43
Weakly Supervised Object Localization as Domain Adaption
Zhu, L., She, Q., Chen, Q., You, Y., Wang, B., & Lu, Y. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.01423 · Cited by 42
“data augmentations [27, 14, 29, 4], novel network structures [4, 30, 1, 31], or post-processes [2, 13].”
BackgroundMethodInfluentialDECAPS: Detail-Oriented Capsule Networks
Mobiny, A., Yuan, P., Cicalese, P. A., & Nguyen, H. V. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-59710-8_15 · Cited by 41
Lee, J., Kim, E., Mok, J., & Yoon, S. (2022) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2022.3166916 · Cited by 46
“…the important, or discriminative, regions of an One popular method of manipulation is erasure, which iteratively removes the discriminative regions which have already been identified, forcing the classifier to find new regions of the target object [23], [24], [25], [28], [29], [30], [32], [33].”
BackgroundGridMix: Strong regularization through local context mapping
Baek, K., Bang, D., & Shim, H. (2020) · Pattern Recognition · doi:10.1016/j.patcog.2020.107594 · Cited by 47
Moniruzzaman, M., Yin, Z., He, Z., Qin, R., & Leu, M. C. (2020) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3394171.3413687 · Cited by 39
“Hide-and-Seek [30] hid random frame sequences to force the network to look at different relevant parts for the temporal action localization task.”
BackgroundLearning representational invariances for data-efficient action recognition
Zou, Y., Choi, J., Wang, Q., & Huang, J. (2022) · Computer Vision and Image Understanding · doi:10.1016/j.cviu.2022.103597 · Cited by 43
Atrous convolutional feature network for weakly supervised semantic segmentation
Xu, L., Xue, H., Bennamoun, M., Boussaïd, F., & Sohel, F. A. (2020) · Neurocomputing · doi:10.1016/j.neucom.2020.09.045 · Cited by 39
Farady, I., Lin, C., & Chang, M. (2023) · Journal of Intelligent Manufacturing · doi:10.1007/s10845-023-02109-0 · Cited by 39
“Unfortunately, not all transformations are useful for every dataset or problem.”
Learning Multi-Modal Class-Specific Tokens for Weakly Supervised Dense Object Localization
Xu, L., Ouyang, W., Bennamoun, M., Boussaïd, F., & Xu, D. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.01877 · Cited by 38
Zhai, W., Wu, P., Zhu, K., Cao, Y. H., Wu, F., & Zha, Z.-J. (2023) · International Journal of Computer Vision · doi:10.1007/s11263-023-01919-2 · Cited by 41
“Extensive exper-Wei”
Zhao, M., Hu, X., Zhang, L., Meng, Q., Chen, Y., & Bruzzone, L. (2024) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2024.3379431 · Cited by 38
“Singh and Lee [35] randomly hide patches of the input image during training, making the model mine more object regions.”
BackgroundZhang, X., Shi, H., Li, C., & Li, P. (2020) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v34i07.6986 · Cited by 37
“Hide-and-seek (Singh and Lee 2017) enforces the model to see different parts of the image and focuses on multiple relevant parts of the object by randomly masking different regions of training images.”
BackgroundZhao, T., Han, J., Yang, L., Wang, B., & Zhang, D. (2021) · International Journal of Computer Vision · doi:10.1007/s11263-021-01473-9 · Cited by 38
Enhancing Mixture-of-Experts by Leveraging Attention for Fine-Grained Recognition
Zhang, L., Huang, S., & Liu, W. (2021) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2021.3117064 · Cited by 39
“We construct input for the next expert using a technique similar to weakly supervised object localization [20], [21], [62].”
MethodCategory-Aware Spatial Constraint for Weakly Supervised Detection
Shen, Y., Ji, R., Yang, K., Deng, C. D., & Wang, C. (2019) · IEEE Transactions on Image Processing · doi:10.1109/tip.2019.2933735 · Cited by 37
“Singh and Lee [15] randomly hided patches in training images and used CAMs to extract bounding boxes in testing images.”
MethodSelf-Supervised Feature Enhancement Networks for Small Object Detection in Noisy Images
Lee, G., Hong, S., & Cho, D. (2021) · IEEE Signal Processing Letters · doi:10.1109/lsp.2021.3081041 · Cited by 36
“It tends to be similar to studies [25] that the deep learning network trained by solving difficult tasks achieves better performance.”
ResultBenchmark Platform for Ultra-Fine-Grained Visual Categorization Beyond Human Performance
Yu, X., Zhao, Y. S., Gao, Y., Yuan, X., & Xiong, S. (2021) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.01012 · Cited by 40
A data augmentation framework by mining structured features for fake face image detection
Guo, Z., Yang, G., Wang, D., & Zhang, D. (2022) · Computer Vision and Image Understanding · doi:10.1016/j.cviu.2022.103587 · Cited by 39
Normalization Matters in Weakly Supervised Object Localization
Kim, J., Choe, J., Yun, S., & Kwak, N. (2021) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.00341 · Cited by 35
“3, especially, CAM, HaS and CutMix).”
InfluentialF-CAM: Full Resolution Class Activation Maps via Guided Parametric Upscaling
Belharbi, S., Sarraf, A., Pedersoli, M., Ayed, I. B., McCaffrey, L. M., & Granger, É. (2022) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv51458.2022.00378 · Cited by 35
“In particular, WSOL methods based on data enhancement [8, 30, 40, 46, 49] aim to encourage the model to be less dependent on most discriminative regions and seek additional regions.”
BackgroundMethodInfluentialLi, J., Li, W., Gertych, A., Knudsen, B. S., Speier, W. F., & Arnold, C. W. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1905.13208 · Cited by 40
“Built upon the attention-based MIL model [20], our model further improves the attention mechanism with instance dropout [40].”
MethodInfluential
Metadata
- Authors
- Singh, K. K., & Lee, Y. J.
- Affiliations
- Intel (United States) (US)
- University of California, Davis (US)
- Venue
- 2017 IEEE International Conference on Computer Vision (ICCV)
- Publisher
- IEEE
- Published
- 2017-10
- Pages
- 3544-3553
- OpenAlex
- W2606520630
- Cited by
- 724
- References
- 57
- Citations by year
- 2026: 19 · 2025: 49 · 2024: 74 · 2023: 121 · 2022: 129 · 2021: 137 · 2020: 98 · 2019: 70 · 2018: 25 · 2017: 2
- FWCI
- 15.59
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- IEEE/CVF International Conference on Computer Vision (ICCV) (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- Nvidia
- Amazon Web Services
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Image and Video Retrieval Techniques
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Visual Attention and Saliency Detection
- Keywords
- weakly supervised object localization, data augmentation, image classification, video action recognition
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
