PCL: Proposal Cluster Learning for Weakly Supervised Object Detection
Open accessThe DOI matches its Crossref record; 65 of the 66 references it cites were found in a registry or an index, and 1 are not checked yet.
Abstract
Weakly Supervised Object Detection (WSOD), using only image-level annotations to train object detectors, is of growing importance in object recognition. In this paper, we propose a novel deep network for WSOD. Unlike previous networks that transfer the object detection problem to an image classification problem using Multiple Instance Learning (MIL), our strategy generates proposal clusters to learn refined instance classifiers by an iterative process. The proposals in the same cluster are spatially adjacent and associated with the same object. This prevents the network from concentrating too much on parts of objects instead of whole objects. We first show that instances can be assigned object or background labels directly based on proposal clusters for instance classifier refinement, and then show that treating each cluster as a small new bag yields fewer ambiguities than the directly assigning label method. The iterative instance classifier refinement is implemented online using multiple streams in convolutional neural networks, where the first is an MIL network and the others are for instance classifier refinement supervised by the preceding one. Experiments are conducted on the PASCAL VOC, ImageNet detection, and MS-COCO benchmarks for WSOD. Results show that our method outperforms the previous state of the art significantly.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
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
“Multi-instance learning is a group of supervised learning algorithms that has seen widespread application in WSOD [202-209].”
From Object Detection in 20 Years: A Survey (2023) · cites this paper as [207]
Citations
Cites
Very Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148 · cited 4×
“Here we train a Fast R-CNN (FRCNN) [7] using the VGG16 model and the same five image scales (horizontal flips only in training).”
MethodInfluentialImageNet 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 · cited 1×
“Recently, due to the development of Convolutional Neural Network (CNN) [1], [2] and the availability of large scale datasets with detailed boundingbox-level annotations [3], [4], [5], there have been great leap forwards in object detection [6], [7], [8], [9], [10], [11].”
BackgroundFaster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
Ren, S., He, K., Girshick, R., & Sun, J. (2017) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2016.2577031 · Cited by 73,472 · cited 1×
“Recently, due to the development of Convolutional Neural Network (CNN) [1], [2] and the availability of large scale datasets with detailed boundingbox-level annotations [3], [4], [5], there have been great leap forwards in object detection [6], [7], [8], [9], [10], [11].”
BackgroundScikit-learn: Machine Learning in Python
Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Müller, A., Nothman, J., Louppe, G., et al. (2012) · arXiv · doi:10.48550/arxiv.1201.0490 · Cited by 71,016 · cited 1×
“The k-means algorithm to produce top ranking proposals is implemented by scikit-learn [56].”
MethodFaster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
Ren, S., He, K., Girshick, R., & Sun, J. (2015) · arXiv · doi:10.48550/arxiv.1506.01497 · Cited by 70,812 · cited 1×
“Recently, due to the development of Convolutional Neural Network (CNN) [1], [2] and the availability of large scale datasets with detailed boundingbox-level annotations [3], [4], [5], there have been great leap forwards in object detection [6], [7], [8], [9], [10], [11].”
BackgroundGradient-based learning applied to document recognition
Lecun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998) · Proceedings of the IEEE · doi:10.1109/5.726791 · Cited by 58,805 · cited 14×
“We finally visualize some success and failure detection results on VOC 2007 trainval by PCL-Ens.+FRCNN, as in Fig.”
BackgroundInfluentialYou Only Look Once: Unified, Real-Time Object Detection
Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.91 · Cited by 53,803 · cited 1×
“Recently, due to the development of Convolutional Neural Network (CNN) [1], [2] and the availability of large scale datasets with detailed boundingbox-level annotations [3], [4], [5], there have been great leap forwards in object detection [6], [7], [8], [9], [10], [11].”
BackgroundMicrosoft COCO: Common Objects in Context
Lin, T.-Y., Maire, M., Belongie, S., Hays, J., Perona, P., Ramanan, D., Dollár, P., & Zitnick, C. L. (2014) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-10602-1_48 · Cited by 45,460
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
(2016) · Cited by 31,590
Self-citation
Cited by
Items 101–200 of 409
Towards Generic Image Manipulation Detection with Weakly-Supervised Self-Consistency Learning
Zhai, Y., Luan, T., Doermann, D., & Yuan, J. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.02046 · Cited by 72
“For example, given image-level categorical labels to predict bounding box [21, 46] or segmentation mask [37, 38, 67] and given video-level categorical labels to predict temporal boundaries of actions [50].”
BackgroundMisclassification in Weakly Supervised Object Detection
Wu, Z., Xu, Y. D., Yang, J. J., & Li, X. (2024) · IEEE Transactions on Image Processing · doi:10.1109/tip.2024.3402981 · Cited by 31
“We integrate the proposed MCC and MCT into the widely-used WSOD framework OICR [10] and its variants [15], [20].”
MethodWeakSAM: Segment Anything Meets Weakly-supervised Instance-level Recognition
Zhu, L., Zhou, J., Liu, Y., Hao, X., Liu, W., & Wang, X. (2024) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3664647.3680960 · Cited by 41
“However, they always need complicated networks and lack high-quality instance segments.”
Self-citationRotated cascade R-CNN: A shape robust detector with coordinate regression
Zhu, Y., Ma, C., & Du, J. (2019) · Pattern Recognition · doi:10.1016/j.patcog.2019.106964 · Cited by 31
“Deep patch learning (DPL) [21] presents a novel method to learn patch features with only image-level annotations and proposal cluster learning (PCL) [22] also trains detector with only image-level annotations by generating proposal clusters for instance classifier refinement.”
Where to Look?: Mining Complementary Image Regions for Weakly Supervised Object Localization
Babar, S., & Das, S. (2021) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv48630.2021.00105 · Cited by 35
“Although supervised object detection methods [21, 23, 24, 27] have made tremendous progress, the fact that they require costly bounding box annotations has led to the exploration of weakly supervised object detection meth-ods [9, 31, 32, 45] using only image-level labels.”
BackgroundA Proposal-based Paradigm for Self-supervised Sound Source Localization in Videos
Xuan, H., Wu, Z., Yang, J. J., Yan, Y., & Alameda-Pineda, X. (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/cvpr52688.2022.00110 · Cited by 40
“The audio information, viewed as a “free” source of weak annotations, is introduced into the traditional WSOD framework to establish associations between visible objects and corresponding audio contents.”
BackgroundMethodInfluentialPoint-Teaching: Weakly Semi-supervised Object Detection with Point Annotations
Ge, Y., Zhou, Q., Wang, X., Shen, C., Wang, Z., & Li, H. (2023) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v37i1.25143 · Cited by 39
Self-citationWang, Y., Wang, L., Lu, H., & He, Y. (2020) · Neurocomputing · doi:10.1016/j.neucom.2020.01.039 · Cited by 31
Gonthier, N., Ladjal, S., & Gousseau, Y. (2021) · Computer Vision and Image Understanding · doi:10.1016/j.cviu.2021.103299 · Cited by 35
“That alone can explain the poor results of PCL on those datasets.”
BackgroundMethodInfluentialZurowietz, M., & Nattkemper, T. W. (2020) · IEEE Access · doi:10.1109/access.2020.3014441 · Cited by 36
“One approach for situations where precise image annotations are costly to acquire is weakly supervised object detection [17]–[19].”
BackgroundUniT: Unified Knowledge Transfer for Any-shot Object Detection and Segmentation
Khandelwal, S., Goyal, R., & Sigal, L. (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/cvpr46437.2021.00589 · Cited by 38
“We evaluate our approach on VOC 2007 [6] which consists of a trainval set of 5011 images for training and 4951 images for test, keeping in line with the prior related works [3, 22, 21, 24, 1].”
MethodHan, P., Ma, C., Jian, C., Chen, L., Bu, S., Xu, S., Zhao, Y., Zhang, C., & Hagino, T. (2022) · Remote Sensing · doi:10.3390/rs14164113 · Cited by 31
Feng, J., Jiang, Q., Zhang, J., Liang, Y., Shang, R., & Jiao, L. (2024) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2024.3403868 · Cited by 31
“Weakly supervised object detection frameworks under point-level annotations [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38] commonly consist of two main stages: box initialization and iterative refinement, as illustrated in Fig.”
BackgroundEnd-to-End Weakly Supervised Object Detection with Sparse Proposal Evolution
Liao, M., Wan, F., Yao, Y., Han, Z., Zou, J., Wang, Y., Feng, B., Yuan, P., & Ye, Q. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-20077-9_13 · Cited by 34
“Proposal cluster learning (PCL) [24,43] further alleviated networks from concentrating on object parts by proposal clustering [43].”
BackgroundSalvage of Supervision in Weakly Supervised Object Detection
Sui, L., Zhang, C.-L., & Wu, J. (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/cvpr52688.2022.01383 · Cited by 34
“PCL [31] clustered proposals to improve pseudo groundtruth mining, and C-MIL [34] improved the MIL loss.”
BackgroundMethodInfluentialCounting trees with point-wise supervised segmentation network
Tong, P., Han, P., Li, S., Li, N., Bu, S., Li, Q., & Li, K. (2021) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2021.104172 · Cited by 30
Activity Driven Weakly Supervised Object Detection
Yang, Z., Mahajan, D., Ghadiyaram, D., Nevatia, R., & Ramanathan, V. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.00303 · Cited by 27
“On Charades, both PCL [41] and our method fails to detect the windows and on HICO-DET, our method fails to localize the kite.”
MethodInfluentialHuo, Y., Qian, X., Li, C., & Wang, W. (2023) · IEEE Geoscience and Remote Sensing Letters · doi:10.1109/lgrs.2023.3283403 · Cited by 27
“More specifically, proposed method surpasses WSDDN [2], OICR [3], PCL [14], MELM [15], DCL [4], PCIR [16], and MIG [17] by 23.”
BackgroundMethodInfluentialWeakly Supervised Open-Vocabulary Object Detection
Lin, J., Shen, Y., Wang, B., Lin, S., Li, K., & Cao, L. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v38i4.28127 · Cited by 34
Multi-scale multi-patch person re-identification with exclusivity regularized softmax
Wang, C., Song, L., Wang, G., Zhang, Q., & Wang, X. (2019) · Neurocomputing · doi:10.1016/j.neucom.2019.11.062 · Cited by 27
“…the effectiveness of multi-scale multi-patch deep feature learning; to further enhance the deep features, some advanced technologies, such as deep ensemble learning [23], weakly-supervised learning [24], neural architecture search [25, 26, 27, 28, 29] and few-shot learning [30], can be adopted.”
BackgroundSelf-citationSpatial Self-Distillation for Object Detection with Inaccurate Bounding Boxes
Wu, D., Chen, P., Yu, X., Li, G., Han, Z., & Jiao, J. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00631 · Cited by 35
“Context information [24, 51], spatial regularization [1, 9, 48], and optimization strategy [44, 48, 46] are proposed to address the problems.”
BackgroundMethodContinuation Multiple Instance Learning for Weakly and Fully Supervised Object Detection
Ye, Q., Wan, F., Liu, C., Huang, Q., & Ji, X. (2021) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2021.3070801 · Cited by 28
“of proposal clusters [31] could prevent networks from concentrating on object parts.”
BackgroundMethodInfluentialCyclic Self-Training With Proposal Weight Modulation for Cross-Supervised Object Detection
Xu, Y., Zhou, C., Yu, X., & Yang, Y. (2023) · IEEE Transactions on Image Processing · doi:10.1109/tip.2023.3261752 · Cited by 27
“Most prior works [7], [8], [9], [10], [11], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29] formulate the weakly-supervised object detection (WSOD) task as a multiple instance learning (MIL) problem.”
Deep multiple instance selection
Li, X., Zhan, D., Yang, J., & Shi, Y. (2021) · Science China Information Sciences · doi:10.1007/s11432-020-3117-3 · Cited by 28
CaT: Weakly Supervised Object Detection with Category Transfer
Cao, T., Du, L., Zhang, X., Chen, S., Zhang, Y., & Wang, Y. (2021) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.00306 · Cited by 30
“[32] use an iterative process, Proposal cluster learning (PCL), to learn refined instance classifiers based on OICR.”
MethodInfluentialActive Learning Strategies for Weakly-Supervised Object Detection
Vo, H. V., Siméoni, O., Gidaris, S., Bursuc, A., Pérez, P., & Ponce, J. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-20056-4_13 · Cited by 37
Weakly Supervised Object Detection for Remote Sensing Images: A Survey
Fasana, C., Pasini, S., Milani, F., & Fraternali, P. (2022) · Remote Sensing · doi:10.3390/rs14215362 · Cited by 29
Learning an Invariant and Equivariant Network for Weakly Supervised Object Detection
Feng, X., Yao, X., Shen, H., Cheng, G., Xiao, B., & Han, J. (2023) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2023.3275142 · Cited by 25
“Thus, some OICR-based approaches focused on exploring contextual information [23], [29], [32], [35], [36], or proposal optimization [37],”
BackgroundReal-time and accurate object detection in compressed video by long short-term feature aggregation
Wang, X., Huang, Z., Liao, B., Huang, L., Gong, Y., & Huang, C. (2021) · Computer Vision and Image Understanding · doi:10.1016/j.cviu.2021.103188 · Cited by 25
“In the future, we would like to explorer effective attention [18] and weakly-supervised learning [31] for video object detection.”
MethodSelf-citationA Region-Based Efficient Network for Accurate Object Detection
Guan, Y., Aamir, M., Hu, Z., Abro, W. A., Rahman, Z.-U., Dayo, Z. A., & Akram, S. (2021) · Traitement du signal · doi:10.18280/ts.380228 · Cited by 25
“Tang et al. [67] designed an object detection model based on proposal cluster learning.”
MethodQian, X., Huo, Y., Cheng, G., Yao, X., Li, K., Ren, H., & Wang, W. (2022) · IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing · doi:10.1109/jstars.2022.3150843 · Cited by 24
Attention Erasing and Instance Sampling for Weakly Supervised Object Detection
Xie, X., Cheng, G., Feng, X., Yao, X., Qian, X., & Han, J. (2023) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2023.3339956 · Cited by 24
“…learning capacity of convolutional neural network (CNN) [35], several researchers proposed end-to-end MIL networks with promising WSOD performances, for example, online instance classifier refinement (OICR) [36], min-entropy latent model (MELM) [37], and proposal cluster learning (PCL) [38].”
BackgroundMethodDiverse Complementary Part Mining for Weakly Supervised Object Localization
Meng, M., Zhang, T., Yang, W., Zhao, J. C., Zhang, Y., & Wu, F. (2022) · IEEE Transactions on Image Processing · doi:10.1109/tip.2022.3145238 · Cited by 27
A weakly supervised tooth‐mark and crack detection method in tongue image
Hui, W., Li, L., Lei, H., Luo, Z., Li, C., & Li, S. (2021) · Concurrency and Computation Practice and Experience · doi:10.1002/cpe.6262 · Cited by 23
Camouflage design, assessment and breaking techniques: a survey
Mondal, A. (2021) · Multimedia Systems · doi:10.1007/s00530-021-00813-6 · Cited by 25
Detection of Ice Hockey Players and Teams via a Two-Phase Cascaded CNN Model
Guo, T., Tao, K., Hu, Q., & Shen, Y. (2020) · IEEE Access · doi:10.1109/access.2020.3033580 · Cited by 26
A Comparative Review of Recent Few-Shot Object Detection AlgorithmsPreprint
Leng, J., Taiyue, C., Gao, X., Yu, Y., Wang, Y., Gao, F., & Wang, Y. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2111.00201 · Cited by 26
Learning to Detect Salient Object with Multi-source Weak Supervision
Zhang, H., Zeng, Y., Lu, H., Zhang, L., Li, J., & Qi, J. (2021) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2021.3059783 · Cited by 22
W2N: Switching from Weak Supervision to Noisy Supervision for Object Detection
Huang, Z., Bao, Y., Dong, B., Zhou, E., & Zuo, W. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-20056-4_41 · Cited by 21
Wang, G., Zhang, X., Peng, Z., Zhang, T., Tang, X., Zhou, H., & Jiao, L. (2024) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2024.3395751 · Cited by 25
“…works have been proposed to improve the detection performance, which consists of designing instance refinement modules [15], [16], [17], [20], [27], [40], introducing spatial prior [12], [29], [41], [42], constructing attention modules [28], [32], and deploying optimization approaches [18],…”
InfluentialBag similarity network for deep multi-instance learning
Wang, X., Yan, Y., Tang, P., Liu, W., & Guo, X. (2019) · Information Sciences · doi:10.1016/j.ins.2019.07.071 · Cited by 21
“Tang et al. [26] proposed a bag-in-bag formulation for modeling contextual information around objects.”
BackgroundSelf-citationDLWL: Improving Detection for Lowshot Classes With Weakly Labelled Data
Ramanathan, V., Wang, R., & Mahajan, D. K. (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/cvpr42600.2020.00936 · Cited by 32
“Following the settings in [47], the WSDDN model used for bootstrapping was trained with MCG [36] proposals for PASCAL VOC07 and selective search [52] proposals for COCO14.”
MethodInfluentialWeakly Supervised Instance Segmentation by Deep Community Learning
Hwang, J., Kim, S., Son, J., & Han, B. (2021) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv48630.2021.00106 · Cited by 22
“Most of existing methods formulate WSOD as Multiple Instance Learning (MIL) problems [11] and attempt to learn detection models via extracting pseudo-ground-truth labels [4, 35, 36, 42].”
BackgroundMethodJia, Q., Wei, S., Ruan, T., Zhao, Y., & Zhao, Y. (2021) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v35i2.16261 · Cited by 24
“All the settings of our four baselines (OICR, PCL, MELM and C-MIL) are kept identical to (Tang et al. 2017; Tang et al. 2020; Wan et al. 2018, 2019) For BGM, both (cid:21) 1 and (cid:21) 2 are set to 0.5.”
MethodInfluentialParallel Detection-and-Segmentation Learning for Weakly Supervised Instance Segmentation
Shen, Y., Cao, L., Chen, Z., Zhang, B., Su, C., Wu, Y., Huang, F., & Ji, R. (2021) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.00809 · Cited by 26
“Other different strategies [38, 39, 40, 41, 42, 43] are also proposed to generate pseudo-ground-truth bounding boxes and assign labels to proposals.”
BackgroundWang, G., Zhang, X., Peng, Z., Tang, X., Zhou, H., & Jiao, L. (2022) · International Joint Conference on Artificial Intelligence · doi:10.24963/ijcai.2022/192 · Cited by 22
“To verify the generalization ability of NDI-WSOD, we insert the NCL mechanism into the classical OICR [Tang et al. , 2017] and PCL [Tang et al. , 2018].”
InfluentialDistilling Knowledge from Refinement in Multiple Instance Detection Networks
Zeni, L. F., & Jung, C. R. (2020) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw50498.2020.00392 · Cited by 20
“Most WSOD methods [2, 1, 19, 5, 18, 23] follow the Multiple Instance Learning (MIL) pipeline [6] to train detectors using only image category level annotations.”
BackgroundMethodInfluentialReal-Time Weakly Supervised Object Detection Using Center-of-Features Localization
Ibrahem, H., Salem, A., & Kang, H. (2021) · IEEE Access · doi:10.1109/access.2021.3064372 · Cited by 23
“…[14] proposed online instance classifier refinement (OICR) to learn the largest available part of an object, but the network could only learn the discriminative parts of objects; later (2018) [15] they proposed a clustering technique to obtain smaller bags of words to refine the objects in an image.”
MethodInformative and Consistent Correspondence Mining for Cross-Domain Weakly Supervised Object Detection
Hou, L., Zhang, Y., Fu, K., & Li, J. (2021) · 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr46437.2021.00980 · Cited by 19
Learning Remote Sensing Object Detection With Single Point Supervision
He, S., Zou, H., Wang, Y., Li, B., Cao, X., & Jing, N. (2023) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2023.3343806 · Cited by 28
“To relieve the burden of box-level labeling, numerous image-level supervised detection methods [11], [12], [13], Authorized licensed use limited to the terms of the applicable license agreement with IEEE.”
BackgroundMethodInfluentialCyclic-Bootstrap Labeling for Weakly Supervised Object Detection
Yin, Y., Deng, J., Zhou, W., Li, L., & Li, H. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00645 · Cited by 19
“We follow the common practice [33, 32] to exploit VGG16 [30] pre-trained on Imagenet [6] as the backbone network, and to apply Selective Search [36] for region proposal generation.”
BackgroundMethodInfluentialOne point is all you need for weakly supervised object detection
Zhang, S., Wang, Z., & Ke, W. (2024) · Pattern Recognition · doi:10.1016/j.patcog.2024.111087 · Cited by 22
Deep learning methods for autonomous driving scene understanding tasks: A review
Awan, M., Whangbo, T. K., & Shin, J. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.128098 · Cited by 24
EHSOD: CAM-Guided End-to-End Hybrid-Supervised Object Detection with Cascade Refinement
Fang, L., Xu, H., Liu, Z., Parisot, S., & Li, Z. (2020) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v34i07.6707 · Cited by 18
Segmentation of acute stroke infarct core using image-level labels on CT-angiography
Giancardo, L., Niktabe, A., Ocasio, L., Abdelkhaleq, R., Salazar‐Marioni, S. A., & Sheth, S. A. (2023) · NeuroImage Clinical · doi:10.1016/j.nicl.2023.103362 · Cited by 20
“…et al., 2016; Lerousseau et al., 2020; Lu et al., 2021; Xu et al., 2019, 2014; Zhang et al., 2022) for histopathology or microscopy images, natural images (Ahn and Kwak, 2018; Huang et al., 2018; Pinheiro and Collobert, 2015; Tang et al., 2020), and 3D brain vessel segmentation (Dang et al., 2022).”
BackgroundDETR with Additional Global Aggregation for Cross-domain Weakly Supervised Object Detection
Tang, Z., Sun, Y., Liu, S., & Yang, Y. (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/cvpr52729.2023.01099 · Cited by 22
Yang, K., Wu, Y., Li, J., Yin, C., & Li, X. (2025) · Knowledge-Based Systems · doi:10.1016/j.knosys.2025.113012 · Cited by 18
PointOBB-v3: Expanding Performance Boundaries of Single Point-Supervised Oriented Object Detection
Zhang, P., Luo, J., Yang, X., Yu, Y., Li, Q., Zhou, Y., Jia, X., Lü, X., Chen, J., Li, X., et al. (2025) · International Journal of Computer Vision · doi:10.1007/s11263-025-02486-4 · Cited by 25
“…(Dietterich et al., 1997; Zhou, 2004) has been widely applied in various weakly supervised tasks like image-supervised (Bilen & Vedaldi, 2016; Tang et al., 2017, 2018; Chen et al., 2020; Shen et al., 2020; Zhang et al., 2021; Feng et al., 2022, 2023) and point-supervised horizontal object…”
Xia, R., Li, G., Huang, Z., Meng, H., & Pang, Y. (2022) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2022.3168547 · Cited by 22
“Different from OICR [15] or PCL [23] which produce purely one or multiple pseudo ground-truth boxes with complex clustering algorithms for Fig.”
BackgroundMethodInfluentialFI-WSOD: Foreground Information Guided Weakly Supervised Object Detection
Yin, Y., Deng, J., Zhou, W., Li, L., & Li, H. (2022) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2022.3198018 · Cited by 18
“To further boost the performance of weakly supervised object detectors, many recent works have been proposed to improve the self-training procedure [11], [17], [18], [19], the MIL strategy [14], [20], and feature representation learning [16], [17].”
BackgroundMethodResultInfluentialQian, X., He, Z., Wang, W., Yue, W., Yao, X., & Cheng, G. (2026) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2026.3651264 · Cited by 22
“Other representative works include PCL [23], TCA [24], PCIR [25], and RINET [26].”
Hua, X., Cui, X., Xu, X., Qiu, S., & Li, Z. (2024) · Pattern Recognition · doi:10.1016/j.patcog.2024.111111 · Cited by 19
Learning to focus: cascaded feature matching network for few-shot image recognition
Chen, M., Wang, X., Luo, H., Geng, Y., & Liu, W. (2021) · Science China Information Sciences · doi:10.1007/s11432-020-2973-7 · Cited by 17
Self-citationAutomatic learning for object detection
Zhang, X., ZHAO, C. F., Luo, H., Zhao, W., Zhong, S., Tang, L., Peng, J., & Jianping Fan, B. (2022) · Neurocomputing · doi:10.1016/j.neucom.2022.02.012 · Cited by 17
Utilizing the Instability in Weakly Supervised Object DetectionPreprint
Gao, Y., Liu, B., Guo, N., Ye, X., Wan, F., You, H., & Fan, D. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1906.06023 · Cited by 17
Guo, C., Ma, Z., Zhao, Y., Cao, C., Jiang, Z., & Zhang, H. (2025) · International Journal of Remote Sensing · doi:10.1080/01431161.2025.2479886 · Cited by 16
“Table 6 shows the comparison of the computational cost of our method with that of OICR and PCL.”
InfluentialWeakly supervised target detection based on spatial attention
Zhao, W., & Xu, L. (2024) · Visual Intelligence · doi:10.1007/s44267-024-00037-y · Cited by 16
“This section compares the proposed method with other representative object detectors for remote sensing images, including OICR [4], WSRPN [5], MELM [6], C-MIL [7], OCR [8], MIST [9], PCL [10], WS-JDS [11], SoS [12], TS2C [20], IM-CFB [24] and CASD [25].”
Dewanto, W., Syarifudin, E., Hidayat, S., & Suryadi, S. (2024) · IJORER International Journal of Recent Educational Research · doi:10.46245/ijorer.v5i2.583 · Cited by 16
“The substandard performance of educators in acquiring knowledge and its subsequent impact on the caliber of education has been examined in several studies, including those conducted by (Firlianza & Suhardi, 2024 Shohib et al., 2024 Tang, 2020 Yuliandra et al., 2024).”
BackgroundPairwise Similarity Knowledge Transfer for Weakly Supervised Object Localization
Rahimi, A. M., Shaban, A., Ajanthan, T., Hartley, R. I., & Boots, B. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-58586-0_24 · Cited by 17
Differentially private regression analysis with dynamic privacy allocation
Pan, K., Gong, M., Feng, K., & Wang, K. (2021) · Knowledge-Based Systems · doi:10.1016/j.knosys.2021.106795 · Cited by 16
Qian, X., Wang, C., Li, C., Li, Z., Zeng, L., Wang, W., & Wu, Q. (2023) · IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing · doi:10.1109/jstars.2023.3304411 · Cited by 15
“%) compared with WSDDN, OICR, PCL, MELM, MIST, DCL, PCIR, MIG, and TCA, respectively.”
BackgroundMethodInfluentialDense center-direction regression for object counting and localization with point supervision
Tabernik, D., Muhovič, J., & Skočaj, D. (2024) · Pattern Recognition · doi:10.1016/j.patcog.2024.110540 · Cited by 16
Toward Joint Thing-and-Stuff Mining for Weakly Supervised Panoptic Segmentation
Shen, Y., Cao, L., Chen, Z., Lian, F., Zhang, B., Su, C., Wu, Y., Huang, F., & Ji, R. (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/cvpr46437.2021.01642 · Cited by 20
“Other different strategies [20, 21, 22, 23, 24, 25, 26] were also proposed to generate pseudo-ground-truth boxes and assign labels to box proposals.”
BackgroundTaghizadeh, M., & Chalechale, A. (2021) · Expert Systems with Applications · doi:10.1016/j.eswa.2021.116105 · Cited by 16
An Automatic Fault Detection Method of Freight Train Images Based on BD-YOLO
Zhang, L., Wang, M., Liu, K., Xiao, M., Wen, Z., & Man, J. (2022) · IEEE Access · doi:10.1109/access.2022.3165835 · Cited by 15
“Tang et al. [29] proposed Weakly Supervised Object Detection Deep Network (WSOD), which learns and improves the classifier in the iterative process by generating proposal clusters.”
BackgroundScaling Novel Object Detection with Weakly Supervised Detection Transformers
LaBonte, T., Song, Y., Wang, X., Vineet, V., & Joshi, N. (2023) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv56688.2023.00017 · Cited by 19
“Subsequent works have improved WSDDN via self-training and box refinement [35, 15, 65, 80, 77, 66, 75, 79, 37, 56], spatial relations [36, 9, 64], optimization methods [71], residual backbones [61], and segmentation [26, 62, 41].”
BackgroundSelf-citationSalvage of Supervision in Weakly Supervised Object Detection and Segmentation
Sui, L., Zhang, C.-L., & Wu, J. (2023) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2023.3243054 · Cited by 18
“In contrast, most recent WSOD methods [11], [12], [13], [14], [15] combined Fast R-CNN with Multiple Instance Learning (MIL), and still use VGG16 [1] as the backbone, while extensive efforts are required to make ResNet work in WSOD [16].”
BackgroundMethodInfluentialMulti-Scale Low-Discriminative Feature Reactivation for Weakly Supervised Object Localization
Wang, B., Yuan, C., Li, B., Ding, X., Li, Z., Wu, Y. N., & Hu, W. (2021) · IEEE Transactions on Image Processing · doi:10.1109/tip.2021.3091833 · Cited by 15
StarNet: towards Weakly Supervised Few-Shot Object Detection
Karlinsky, L., Shtok, J., Alfassy, A., Lichtenstein, M., Harary, S., Schwartz, E., Doveh, S., Sattigeri, P., Feris, R., Bronstein, A., et al. (2021) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v35i2.16268 · Cited by 15
“Third baseline is PCL (Tang et al. 2018a) - recent (non few-shot) WSOD method.”
BackgroundMethodInfluentialEntity-Graph Enhanced Cross-Modal Pretraining for Instance-level Product Retrieval
Dong, X. M., Zhan, X., Wei, Y., Wei, X.-Y., Wang, Y., Lu, M., Cao, X., & Liang, X. (2023) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2023.3291237 · Cited by 16
“PCL[51]buildsproposalclustersinaniterative mannertofacilitatetheconstructionofinstanceclassifiers.”
BackgroundProgressive Frame-Proposal Mining for Weakly Supervised Video Object Detection
Han, M., Wang, Y., Li, M., Chang, X., Yang, Y., & Qiao, Y. (2024) · IEEE Transactions on Image Processing · doi:10.1109/tip.2024.3364536 · Cited by 16
“As mentioned before, most previous approaches [13], [14] perform Selective Search [38] to generate object proposals of each image in an unsupervised manner.”
BackgroundMethodInfluentialZheng, S., Wu, Z., Xu, Y., & Wei, Z. (2024) · Remote Sensing · doi:10.3390/rs16071203 · Cited by 14
Weakly Supervised Instance Segmentation Based on Two-Stage Transfer Learning
Sun, Y., Liao, S., Gao, C., Xie, C., Yang, F., Zhao, Y., & Sagata, A. (2020) · IEEE Access · doi:10.1109/access.2020.2969480 · Cited by 14
Weakly Supervised Object Detection With Class Prototypical Network
Li, H., Li, Y., Cao, Y., Han, Y., Jin, Y., & Wei, Y. (2022) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2022.3187257 · Cited by 15
“They exploit contextual information [9], [17]–[20], additional segmentation maps [21]–[26], re-finement modules [14], [15], [27]–[29], cross-image information [8] and other information [30] to refine the output of WS-DDN.”
BackgroundMethodInfluentialUncertainty-Aware Graph-Guided Weakly Supervised Object Detection
Zhu, Y., Zhang, Y., Ding, M., & Zuo, W. (2022) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2022.3232487 · Cited by 15
“Based on PCL [22], Chen et al. [24] proposed a spatial likelihood voting (SLV) module to converge the proposal localizing process.”
MethodSelf-Supervised Object Detection and Retrieval Using Unlabeled Videos
Amrani, E., Ben‐Ari, R., Shapira, I., Hakim, T., & Bronstein, A. M. (2020) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw50498.2020.00485 · Cited by 15
“Although huge progress has been made in the field, with the recent weakly-supervised training methods [3, 8, 23, 30, 25] coming close to results achieved in fully-supervised approaches [17, 10], the size, quality, and availability of labeled data are currently becoming a major bottleneck.”
BackgroundMethodResultInfluentialMany-Shot from Low-Shot: Learning to Annotate Using Mixed Supervision for Object Detection
Biffi, C., McDonagh, S., Torr, P. H. S., Leonardis, A., & Parisot, S. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-58598-3_3 · Cited by 14
Li, S., Xue, L., Feng, L., Yao, C., & Wang, D. (2022) · Computers & Electrical Engineering · doi:10.1016/j.compeleceng.2022.108154 · Cited by 14
Refining and reweighting pseudo labels for weakly supervised object detection
Feng, Y., Zeng, H., Li, S., Liu, Q., & Wang, Y. (2024) · Neurocomputing · doi:10.1016/j.neucom.2024.127387 · Cited by 13
Xuan, H., Wu, Z., Yang, J. J., Jiang, B., Luo, L., Alameda-Pineda, X., & Yan, Y. (2024) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2024.3363508 · Cited by 13
“Our three-stream framework, like other end-to-end WSOD [26], [27], [28], is constructed on the top of the Region of Interest (RoI) pooling layer [58], RoI align layer [59] or Spatial Pyramid Pooling (SPP) layer [60] to distribute convolutional (Conv) computations for model acceleration.”
BackgroundMethodQian, X., Lin, C., Chen, Z., & Wang, W. (2024) · Remote Sensing · doi:10.3390/rs16091532 · Cited by 13
Weakly supervised easy-to-hard learning for object detection in image sequences
Yu, H., Guo, D., Yan, Z., Fu, L., Simmons, J. P., Przybyla, C., & Wang, S. (2020) · Neurocomputing · doi:10.1016/j.neucom.2020.02.075 · Cited by 14
“In the area of object detection, weak supervision was sometimes provided to avoid manually labeling the object locations, e.g., only using the image-level labels [30,31] to discover the object locations where the Multiple Instance Learning (MIL) based methods are effective.”
MethodFCC-Net: A Full-Coverage Collaborative Network for Weakly Supervised Remote Sensing Object Detection
Chen, S., Shao, D., Shu, X., Zhang, C., & Wang, J. (2020) · Electronics · doi:10.3390/electronics9091356 · Cited by 14
Weakly supervised foreground learning for weakly supervised localization and detection
Zhang, C.-L., Li, Y., & Wu, J. (2022) · Pattern Recognition · doi:10.1016/j.patcog.2022.109279 · Cited by 13
Kim, D., Lee, G., Jeong, J., & Kwak, N. (2020) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v34i07.6784 · Cited by 16
Self-Supervised Object Detection via Generative Image Synthesis
Mustikovela, S. K., Mello, S. D., Prakash, A., Iqbal, U., Liu, S., Nguyen-Phuoc, T., Rother, C., & Kautz, J. (2021) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.00849 · Cited by 15
“Compared to Wetectron (mAP of 38.1 for All) and PCL (mAP of 33.2 for All), SSOD (mAP of 68.4 for All) has ∼2X better detection accuracy.”
MethodInfluentialWu, X., Li, M., Geng, Y., & Qian, H. (2024) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2024.3522658 · Cited by 14
“In the realm of computer vision, UL methods predicated on clustering [27] and dimension reduction [28] are introduced.”
Yang, L., Wang, S., Wang, X., Wang, S., Lu, Y., & Wang, B. (2025) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2025.112110 · Cited by 12
Self-citationA blockchain‐enabled learning model based on distributed deep learning architecture
Zhang, Y., Liang, Y., Jia, B., Wang, P., & Zhang, X. (2022) · International Journal of Intelligent Systems · doi:10.1002/int.22907 · Cited by 13
Point Beyond Class: A Benchmark for Weakly Semi-supervised Abnormality Localization in Chest X-Rays
Ji, H., Liu, H., Li, Y., Xie, J., He, N., Huang, Y., Wei, D., Chen, X., Shen, L., & Zheng, Y. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-16437-8_24 · Cited by 12
Metadata
- Authors
- Tang, P., Wang, X., Bai, S., Shen, W., Bai, X., Liu, W., & Yuille, A.
- Affiliations
- Huazhong University of Science and Technology (CN)
- University of Oxford (GB)
- Shanghai University (CN)
- Johns Hopkins University (US)
- Venue
- IEEE Transactions on Pattern Analysis and Machine Intelligence
- Published
- 2020-01-01
- Volume
- 42
- Issue
- 1
- Pages
- 176-191
- ISSN
- 0162-8828 (ISSN-L)
- 2160-9292
- 1939-3539
- OpenAlex
- W2813911573
- Cited by
- 406
- References
- 62
- Citations by year
- 2026: 16 · 2025: 41 · 2024: 54 · 2023: 59 · 2022: 74 · 2021: 82 · 2020: 50
- FWCI
- 20.27
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- Locations
- IEEE Transactions on Pattern Analysis and Machine Intelligence (publishedVersion)
- PubMed (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- National Natural Science Foundation of China (61733007, 61572207, 61876212, 61672336)
- ONR (N00014-15-1-2356)
- Hubei Scientific and Technical Innovation Key Project
- HUST Academic Frontier Youth Team
- 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 detection, multiple instance learning, image-level annotations, iterative refinement, multi-stream convolutional neural network
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
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.