What Makes for Effective Detection Proposals?
Open accessThe DOI matches its Crossref record; 69 of the 69 references it cites were found in a registry or an index.
Abstract
Current top performing object detectors employ detection proposals to guide the search for objects, thereby avoiding exhaustive sliding window search across images. Despite the popularity and widespread use of detection proposals, it is unclear which trade-offs are made when using them during object detection. We provide an in-depth analysis of twelve proposal methods along with four baselines regarding proposal repeatability, ground truth annotation recall on PASCAL, ImageNet, and MS COCO, and their impact on DPM, R-CNN, and Fast R-CNN detection performance. Our analysis shows that for object detection improving proposal localisation accuracy is as important as improving recall. We introduce a novel metric, the average recall (AR), which rewards both high recall and good localisation and correlates surprisingly well with detection performance. Our findings show common strengths and weaknesses of existing methods, and provide insights and metrics for selecting and tuning proposal methods.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
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?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“It is noteworthy that the Recall-to-IoU metric is just loosely [19], [20], [21] related to the ultimate detection accuracy.”
From Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks (2017) · cites this paper as [20]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“In the case of sliding window based detectors, the imbalance between backgrounds and objects could be as extreme as 10 7 :1 [71].”
From Object Detection in 20 Years: A Survey (2023) · cites this paper as [71]
Experimental — accuracy not yet measured
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
Faster 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×
“Indeed, there has been some very recent work in this space [60], [61], [62] that shows promising”
BackgroundFaster 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×
“Indeed, there has been some very recent work in this space [60], [61], [62] that shows promising”
BackgroundImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333
Microsoft 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
Self-citationRich 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
Girshick, R. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.169 · Cited by 28,105
Robust Real-Time Face Detection
Viola, P., & Jones, M. J. (2004) · International Journal of Computer Vision · doi:10.1023/b:visi.0000013087.49260.fb · Cited by 14,097
Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2015) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2015.2389824 · Cited by 11,971
Object Detection with Discriminatively Trained Part-Based Models
Felzenszwalb, P. F., Girshick, R. B., McAllester, D., & Ramanan, D. (2010) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2009.167 · Cited by 9,853
The Pascal Visual Object Classes Challenge: A Retrospective
Everingham, M., Eslami, S. M. A., Van Gool, L., Williams, C. K. I., Winn, J., & Zisserman, A. (2015) · International Journal of Computer Vision · doi:10.1007/s11263-014-0733-5 · Cited by 7,139 · cited 3×
“We test the generalisation of proposal methods by evaluating on datasets with many additional classes in §4.”
Method
Research works printed but not linked: 7
- [28] A. Humayun, F. Li, and J. M. Rehg, 'Rigor: Recycling inference in graph cuts for generating object regions,' in CVPR , 2014. -
- [22] P. Krähenbühl and V. Koltun, 'Geodesic object proposals,' in ECCV , 2014. -
- [51] P. Dollár, R. Appel, and W. Kienzle, 'Crosstalk cascades for framerate pedestrian detection,' in ECCV , 2012. -
- [59] K. Tang, A. Joulin, L.-J. Li, and L. Fei-Fei, 'Co-localization in realworld images,' in CVPR , 2014. -
- [57] S. Vicente, C. Rother, and V. Kolmogorov, 'Object cosegmentation,' in CVPR , 2011. -
- [54] N. Chavali, H. Agrawal, A. Mahendru, and D. Batra, 'Objectproposal evaluation protocol is 'gameable',' arXiv:1505.05836 , 2015. - · arXiv
- szegedy, 0, Scalable, high-quality object detection, arXiv 1412 1441
Cited by
Items 101–200 of 764
Oriented object detection in optical remote sensing images using deep learning: a survey
Wang, K., Wang, Z., Li, Z., Su, A., Teng, X., Pan, E., Liu, M., & Yu, Q. (2025) · Artificial Intelligence Review · doi:10.1007/s10462-025-11256-0 · Cited by 118
“In the first stage, a sparse set of high-quality region proposals, that can potentially contain objects, are generated via RPN (Chavali et al., 2016; Hosang et al., 2016).”
Segmentation of Lung Nodule in CT Images Based on Mask R-CNN
Liu, M., Dong, J., Dong, X., Yu, H. L., & Qi, L. (2018) · International Conference on Awareness Science and Technology (iCAST) · doi:10.1109/icawst.2018.8517248 · Cited by 113
“The first step in the R-CNN approach was to generate a manageable number of candidate object regions [16,17].”
MethodPedestrian and Cyclist Detection and Intent Estimation for Autonomous Vehicles: A Survey
Ahmed, S., Huda, M. N., Rajbhandari, S., Saha, C., Elshaw, M., & Kanarachos, S. A. (2019) · Applied Sciences · doi:10.3390/app9112335 · Cited by 112
“A comparison of the DL architectures can be found in [100] and summarised in Table 1.”
BackgroundMulti-modal deep feature learning for RGB-D object detection
Xu, X., Li, Y., Wu, G., & Luo, J. (2017) · Pattern Recognition · doi:10.1016/j.patcog.2017.07.026 · Cited by 109
Dental enumeration and multiple treatment detection on panoramic X-rays using deep learning
Yüksel, A. E., Gültekin, S., Simsar, E., Özdemir, Ş. D., Gündoğar, M., Tokgöz, S. B., & Hamamcı, İ. E. (2021) · Scientific Reports · doi:10.1038/s41598-021-90386-1 · Cited by 107
SCC: Semantic Context Cascade for Efficient Action Detection
Heilbron, F. C., Barrios, W., Escorcia, V. A., & Ghanem, B. (2017) · 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/cvpr.2017.338 · Cited by 112
“Following the success and ubiquity of 2D object proposals in 2D object detectors [18], spatiotemporal proposals [43, 48, 53] and temporal activity proposals [5, 9, 30, 37] have emerged as a key pre-processing step to avoid the exhaustive sliding window approach for action detection [8, 10, 32].”
BackgroundMethodSimultaneously Detecting and Counting Dense Vehicles From Drone Images
Li, W., Li, H., Wu, Q., Chen, X., & Ngan, K. N. (2019) · IEEE Transactions on Industrial Electronics · doi:10.1109/tie.2019.2899548 · Cited by 96
“Compared with the average recall, these metrics are more suitable for the detection task, while considering the recall and false detection rate [44], [45].”
Li, L., Zhou, Z., Wang, B., Miao, L., & Zong, H. (2020) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2020.2995477 · Cited by 101
“In addition, we use the average recall (AR) [42] to quanti-”
MethodAdaptive Region Proposal With Channel Regularization for Robust Object Tracking
Lu, X., Ma, C., Ni, B., & Yang, X. (2019) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2019.2944654 · Cited by 96
“Region proposal methods [46]–[51] typically rely on inexpensive features and economical inference schemes to provide candidate bounding boxes for object detection and recognition tasks.”
BackgroundMethodTowards Real-Time Object Detection on Embedded Systems
Mao, H., Yao, S., Tang, T., Li, B., Yao, J., & Wang, Y. (2018) · IEEE Transactions on Emerging Topics in Computing · doi:10.1109/tetc.2016.2593643 · Cited by 94
“What Makes for Effective Detection Proposals? [28].”
InfluentialVehicle license plate detection using region-based convolutional neural networks
Rafique, M. A., Pedrycz, W., & Jeon, M. (2017) · Soft Computing · doi:10.1007/s00500-017-2696-2 · Cited by 99
“There are various techniques to compute the region proposals discussed in detail in Chavali et al. (2015) and Hosang et al. (2016).”
MethodZoom Out-and-In Network with Map Attention Decision for Region Proposal and Object Detection
Li, H., Liu, Y., Ouyang, W., & Wang, X. (2019) · International Journal of Computer Vision · doi:10.1007/s11263-018-1101-7 · Cited by 90
“AR summarizes the general proposal performance and is shown to correlate with the average precision (AP) performance of a detector better than other metrics (Hosang et al., 2015).”
BackgroundTemporal Action Detection with Structured Segment Networks
Zhao, Y., Xiong, Y., Wang, L., Wu, Z., Tang, X., & Lin, D. (2019) · International Journal of Computer Vision · doi:10.1007/s11263-019-01211-2 · Cited by 102
Real-Time Water Surface Object Detection Based on Improved Faster R-CNN
Zhang, L., Zhang, Y., Zhang, Z., Shen, J., & Wang, H. (2019) · Sensors · doi:10.3390/s19163523 · Cited by 95
“The Recall–IoU metric can be used to evaluate the localization performance, which is related to the final detection accuracy [36].”
MethodOpen Plant Phenotype Database of Common Weeds in Denmark
Madsen, S. L., Mathiassen, S. K., Dyrmann, M., Laursen, M. S., Paz, L.-C., & Jørgensen, R. N. (2020) · Remote Sensing · doi:10.3390/rs12081246 · Cited by 96
“AR is the maximum recall given a fixed number of detections per image as defined by Hosang et al. [40]: where IoU ( gt i ) is the IoU between the annotation gt i and the closest detection proposal, and n is the number of detections per image [40].”
MethodSewer pipe defect detection via deep learning with local and global feature fusion
Li, D., Xie, Q., Yu, Z., Wu, Q., Zhou, J., & Wang, J. (2021) · Automation in Construction · doi:10.1016/j.autcon.2021.103823 · Cited by 95
Robust Vehicle Detection in Aerial Images Based on Cascaded Convolutional Neural Networks
Zhong, J., Lei, T., & Yao, G. (2017) · Sensors · doi:10.3390/s17122720 · Cited by 86
Beyond Object Proposals: Random Crop Pooling for Multi-Label Image Recognition
Wang, M., Luo, C., Hong, R., Tang, J., & Feng, J. (2016) · IEEE Transactions on Image Processing · doi:10.1109/tip.2016.2612829 · Cited by 97
“The positive training examples and negative training examples are determined by thresholding the overlapping ratio with the ground truth bounding boxes.”
MethodDeep Semantic Feature Matching
Ufer, N., & Ommer, B. (2017) · 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/cvpr.2017.628 · Cited by 91
“[20] utilize modern object proposal methods [4, 36, 45, 51, 24].”
Du, X., Ang, M. H., & Rus, D. L. (2017) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2017.8202234 · Cited by 94
“Metrics: To evaluate the proposal network, following [9], [13] and [14], the Oracle recall and average recall (AR) [27] are reported, which are highly correlated to the final detection results.”
MethodMFR-CNN: Incorporating Multi-Scale Features and Global Information for Traffic Object Detection
Zhang, H., Wang, K., Tian, Y., Gou, C., & Wang, F. (2018) · IEEE Transactions on Vehicular Technology · doi:10.1109/tvt.2018.2843394 · Cited by 87
“We first compute the Recall-to-IoU metric [56] to evaluate the performance of RPN.”
MethodA Convnet for Non-maximum Suppression
Hosang, J., Benenson, R., & Schiele, B. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-45886-1_16 · Cited by 80
“An oracle analysis of “perfect NMS” shows potential improvements from 68% to 75% mAP, accounting for almost a quarter of the remaining Pascal VOC detection problem (Hosang et al., 2015).”
BackgroundMethodSelf-citationCollaborative Deep Reinforcement Learning for Joint Object Search
Kong, X., Xin, B., Wang, Y., & Hua, G. (2017) · 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/cvpr.2017.748 · Cited by 94
Multiclass Weighted Loss for Instance Segmentation of Cluttered Cells
Peña, F. A. G., Fernández, P. D. M., Ren, T. I., Yui, M. A., Rothenberg, E. V., & Cunha, A. (2018) · Proceedings - International Conference on Image Processing · doi:10.1109/icip.2018.8451187 · Cited by 84
“Contrary to the Intersection over Union (IoU) metric for detection [17] which uses bounding boxes, the Jaccard Index calculates the instance adequacy from object segmentation.”
MethodPham, C. C., & Jeon, J. W. (2017) · Signal Processing Image Communication · doi:10.1016/j.image.2017.02.007 · Cited by 94
High-Quality R-CNN Object Detection Using Multi-Path Detection Calibration Network
Chen, X., Li, H., Wu, Q., Ngan, K. N., & Xu, L. (2020) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2020.2987465 · Cited by 91
“WE GIVE TWO EVALUATION METRICS: MEAN AVERAGE PRECISION (MAP) AND MEAN AVERAGE RECALL (MAR) [40]”
BackgroundImproving object proposals with multi-thresholding straddling expansion
Chen, X., Ma, H., Wang, X., & Zhao, Z. (2015) · 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/cvpr.2015.7298874 · Cited by 83
Object detection through search with a foveated visual system
Akbaş, E., & Eckstein, M. P. (2017) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1005743 · Cited by 91
Krahenbuhl, P., & Koltun, V. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298765 · Cited by 73
Identification and enumeration of cyanobacteria species using a deep neural network
Baek, S., Pyo, J., Pachepsky, Y. A., Park, Y., Ligaray, M. V., Ahn, C., Kim, Y.-H., Chun, J. A., & Cho, K. H. (2020) · Ecological Indicators · doi:10.1016/j.ecolind.2020.106395 · Cited by 85
BING: Binarized normed gradients for objectness estimation at 300fps
Cheng, M., Liu, Y., Lin, W.-Y., Zhang, Z., Rosin, P. L., & Torr, P. H. S. (2018) · Computational Visual Media · doi:10.1007/s41095-018-0120-1 · Cited by 82
Computer vision applications in offsite construction
Alsakka, F., Assaf, S., El-chami, I., & Al‐Hussein, M. (2023) · Automation in Construction · doi:10.1016/j.autcon.2023.104980 · Cited by 88
Deepblueberry: Quantification of Blueberries in the Wild Using Instance Segmentation
Gonzalez, S., Arellano, C., & Tapia, J. E. (2019) · IEEE Access · doi:10.1109/access.2019.2933062 · Cited by 85
“Most methods relying on R-CNN use segment proposals to achieve instance segmentation [12]–[14], [17].”
MethodContactless Palmprint Identification Using Deeply Learned Residual Features
Liu, Y., & Kumar, A. (2020) · IEEE Transactions on Biometrics Behavior and Identity Science · doi:10.1109/tbiom.2020.2967073 · Cited by 87
Human action recognition approaches with video datasets—A survey
Özyer, T., Selin, A. D., & Alhajj, R. (2021) · Knowledge-Based Systems · doi:10.1016/j.knosys.2021.106995 · Cited by 81
Ham, B., Cho, M., Schmid, C., & Ponce, J. (2016) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · Cited by 73
Dairy goat detection based on Faster R-CNN from surveillance video
Wang, D., Tang, J., Zhu, W., Li, H., Xin, J. L., & He, D. (2018) · Computers and Electronics in Agriculture · doi:10.1016/j.compag.2018.09.030 · Cited by 78
Rao, J., Qiao, Y., Ren, F., Wang, J., & Du, Q. (2017) · Sensors · doi:10.3390/s17091951 · Cited by 72
“Compared with the traditional sliding window algorithm, the region proposal method (also called the detection proposal method) [58] is a better solution for region selection.”
MethodXia, L., Cui, J., Shen, R., Xu, X. W., Gao, Y., & Li, X. (2020) · International Journal of Advanced Robotic Systems · doi:10.1177/1729881420919185 · Cited by 76
“In principle, R-CNN generates the region proposals via a selective search, 29 and the feature extraction and classification are, respectively, achieved via AlexNet 30 and support vector machines (SVMs).”
Nguyen-Ha, P., Arsalan, M., Koo, J. H., Naqvi, R. A., Truong, N. Q., & Park, K. R. (2018) · Sensors · doi:10.3390/s18061703 · Cited by 81
A deep learning approach for anthracnose infected trees classification in walnut orchards
Anagnostis, A., Tagarakis, A. C., Asiminari, G., Papageorgiou, E. I., Katerıs, D., Moshou, D., & Bochtis, D. (2021) · Computers and Electronics in Agriculture · doi:10.1016/j.compag.2021.105998 · Cited by 74
Pose-RCNN: Joint object detection and pose estimation using 3D object proposals
Braun, M., Rao, Q., Wang, Y., & Flohr, F. B. (2016) · IEEE International Conference on Intelligent Transportation Systems (ITSC) · doi:10.1109/itsc.2016.7795763 · Cited by 72
“Average recall (AR) defined in [11] is used as the metric for comparison.”
BackgroundMethodDetecting Small Signs from Large Images
Meng, Z., Fan, X., Chen, X., Chen, M., & Tong, Y. (2017) · IEEE International Conference on Information Reuse and Integration (IRI) · doi:10.1109/iri.2017.57 · Cited by 72
“, candidate regions for object detection, using an object proposal algorithm [13], produces features for each proposal region using a CNN, and makes final predictions using an SVM.”
MethodObject-Proposal Evaluation Protocol is ‘Gameable’
Chavali, N., Agrawal, H., Mahendru, A., & Batra, D. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.97 · Cited by 65
“In [59], authors present an analysis of various proposal methods regarding proposal repeatability, ground truth annotation recall, and their impact on detection performance.”
BackgroundA Fully Spiking Hybrid Neural Network for Energy-Efficient Object Detection
Chakraborty, B., She, X., & Mukhopadhyay, S. (2021) · IEEE Transactions on Image Processing · doi:10.1109/tip.2021.3122092 · Cited by 77
“To evaluate the quality of the proposals, we use the standard average recall (AR@k) [43].”
MethodAutomatic damage detection using anchor-free method and unmanned surface vessel
He, Z., Jiang, S., Zhang, J., & Wu, G. (2021) · Automation in Construction · doi:10.1016/j.autcon.2021.104017 · Cited by 65
Deep Regression Neural Network for Industrial Surface Defect Detection
He, Z., & Liu, Q. (2020) · IEEE Access · doi:10.1109/access.2020.2975030 · Cited by 68
“(3) [47]. where n is the number of defects.”
BackgroundGan, H., Ou, M., Huang, E., Xu, C., Li, S., Li, J., Liu, K., & Xue, Y. (2021) · Computers and Electronics in Agriculture · doi:10.1016/j.compag.2021.106357 · Cited by 63
Li, Y., Fu, K., Sun, H., & Sun, X. (2018) · Remote Sensing · doi:10.3390/rs10020243 · Cited by 61
“Many object proposal methods were proposed and compared in the past few years [26], such as Bing [27], SelectiveSearch [28], and EdgeBoxes [29].”
MethodLarge scale detailed mapping of dengue vector breeding sites using street view images
Haddawy, P., Wettayakorn, P., Nonthaleerak, B., Yin, M. S., Wiratsudakul, A., Schöning, J., Laosiritaworn, Y., Balla, K., Euaungkanakul, S., Quengdaeng, P., et al. (2019) · PLoS neglected tropical diseases · doi:10.1371/journal.pntd.0007555 · Cited by 66
Garanderie, G. P. D. L., Atapour–Abarghouei, A., & Breckon, T. P. (2018) · Lecture notes in computer science · doi:10.1007/978-3-030-01261-8_48 · Cited by 65
Multistage Object Detection With Group Recursive Learning
Li, J., Liang, X., Li, J., Wei, Y., Xu, T., Feng, J., & Yan, S. (2018) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2017.2772796 · Cited by 59
Persistent Aerial Tracking system for UAVs
Mueller, M. F., Sharma, G., Smith, N., & Ghanem, B. (2016) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2016.7759253 · Cited by 63
Wang, G., Guo, J.-M., Chen, Y., Li, Y., & Xu, Q. (2019) · IEEE Access · doi:10.1109/access.2019.2897283 · Cited by 61
“Moreover, the deep learningmethods with high quality region proposals [17], [18] have achieved excellent localization and classification accuracy.”
BackgroundCascade region proposal and global context for deep object detection
Zhong, Q., Li, C., Zhang, Y., Xie, D., Yang, S., & Pu, S. (2019) · Neurocomputing · doi:10.1016/j.neucom.2017.12.070 · Cited by 61
Li, W., Hsu, C.-Y., & Hu, M. (2021) · Annals of the American Association of Geographers · doi:10.1080/24694452.2021.1877527 · Cited by 52
“One pioneer research belonging to this second kind is the work by Tang et al. [38], which propose a multi-stage region proposal network where each stage filters out part of the original proposals based on different strategies.”
ComNet: Combinational Neural Network for Object Detection in UAV-Borne Thermal Images
Li, M.-L., Zhao, X., Li, J.-S., & Nan, L. (2020) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2020.3029945 · Cited by 43
“The R-CNN approach [19] is first to get a manageable number of candidate object regions [20], [21] and evaluate convolutional networks independently on each region of interest (ROI).”
MethodBoosting Object Proposals: From Pascal to COCO
Pont-Tuset, J., & Gool, L. V. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.181 · Cited by 52
Yu, S.-L., Westfechtel, T., Hamada, R., Ohno, K., & Tadokoro, S. (2017) · 2017 IEEE International Symposium on Safety, Security and Rescue Robotics (SSRR) · doi:10.1109/ssrr.2017.8088147 · Cited by 53
“Since sliding a detection window over the whole image for classification is computationally inefficient, the most popular choice is to adopt a two stage detection framework [7],[8].”
BackgroundMethodSalient Object Detection via Fuzzy Theory and Object-Level Enhancement
Zhou, Y., Mao, A., Huo, S., Lei, J., & Kung, S. (2018) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2018.2845667 · Cited by 56
“An object proposal can be represented as two forms [29]: a sliding window, and a grouping region.”
BackgroundPon, A., Adrienko, O., Harakeh, A., & Waslander, S. L. (2018) · Canadian Conference on Computer and Robot Vision · doi:10.1109/crv.2018.00024 · Cited by 55
A Comprehensive Review of Modern Object Segmentation Approaches
Wang, Y., Ahsan, U., Li, H., & Hagen, M. (2022) · Foundations and Trends® in Computer Graphics and Vision · doi:10.1561/0600000097 · Cited by 65
Pedestrian Detection Based on Fast R-CNN and Batch Normalization
Zhao, Z., Bian, H., Hu, D., Cheng, W., & Glotin, H. (2017) · Lecture notes in computer science · doi:10.1007/978-3-319-63309-1_65 · Cited by 55
Using Deep Learning for Collecting Data about Museum Visitor Behavior
Ferrato, A., Limongelli, C., Mezzini, M., & Sansonetti, G. (2022) · Applied Sciences · doi:10.3390/app12020533 · Cited by 65
Multitask Learning for Object Localization With Deep Reinforcement Learning
Wang, Y., Zhang, L., Wang, L., & Wang, Z. (2018) · IEEE Transactions on Cognitive and Developmental Systems · doi:10.1109/tcds.2018.2885813 · Cited by 54
“We followed the evaluation methods in [42] and compared with the active detection model proposed by Caicedo and Lazebnik [10] and the edge-box”
MethodA multispectral feature fusion network for robust pedestrian detection
Song, X., Gao, S., & Chen, C. (2020) · Alexandria Engineering Journal · doi:10.1016/j.aej.2020.05.035 · Cited by 53
Weakly Supervised Object Boundaries
Khoreva, A., Benenson, R., Omran, M., Hein, M., & Schiele, B. (2016) · 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/cvpr.2016.27 · Cited by 56
“It is an enabling ingredient for many vision tasks such as image/video segmentation [1, 10], object proposals [15], object detection [35], and semantic labelling [2].”
BackgroundSelf-citationMultiple-Oriented and Small Object Detection with Convolutional Neural Networks for Aerial Image
Chen, C., Zhong, J., & Tan, Y. (2019) · Remote Sensing · doi:10.3390/rs11182176 · Cited by 54
Embedded Real-Time Vehicle and Pedestrian Detection Using a Compressed Tiny YOLO v3 Architecture
Falaschetti, L., Manoni, L., Palma, L., Pierleoni, P., & Turchetti, C. (2024) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2024.3447453 · Cited by 56
“More recently several methods that do not use sliding windows have been proposed [9], and they can be divided into two main classes: grouping methods and window scoring methods.”
Image Co-localization by Mimicking a Good Detector’s Confidence Score Distribution
Li, Y., Liu, L., Shen, C., & van den Hengel, A. (2016) · Lecture notes in computer science · doi:10.1007/978-3-319-46475-6_2 · Cited by 59
Sequential Optimization for Efficient High-Quality Object Proposal Generation
Zhang, Z., Liu, Y., Chen, X., Zhu, Y., Cheng, M., Saligrama, V., & Torr, P. H. S. (2017) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2017.2707492 · Cited by 52
“As BING++ predicts a small set of object bounding boxes based on edge and segmentation information, it may suffer similar limitations as other edge or segmentation based proposal algorithms (see proposal repeatability section in [37]).”
InfluentialAccurate object classification and detection by faster-RCNN
Lokanath, M., Kumar, K. S., & Keerthi, E. S. (2017) · IOP Conference Series Materials Science and Engineering · doi:10.1088/1757-899x/263/5/052028 · Cited by 51
“When training [12] the RPN we consider only 256 anchors such that the positive and negative labels are almost equal and calculate the loss function.”
MethodTableSense: Spreadsheet Table Detection with Convolutional Neural Networks
Dong, H., Liu, S., Han, S., Fu, Z., & Zhang, D. (2019) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v33i01.330169 · Cited by 50
“Previous research on table detection has mainly targeted other medias, e.g. HTML (Wang et al. 2012; Zhai and Liu 2005; Wang and Hu 2002), images (Gatos et al. 2005; Zuyev 1997; Hu et al. 1999; Liu, Mitra, and Giles 2008; Shafait and Smith 2010) and PDFs (Fang et al. 2012; Liu et al. 2007).”
MethodLogo Recognition Using CNN Features
Bianco, S., Buzzelli, M., Mazzini, D., & Schettini, R. (2015) · Lecture notes in computer science · doi:10.1007/978-3-319-23234-8_41 · Cited by 49
“[18] introduced a class agnostic metric to evaluate the effectiveness of an object proposal algorithm: the Recall versus IoU (Intersection over Union).”
BackgroundA novel target detection method for SAR images based on shadow proposal and saliency analysis
Gao, F., You, J., Wang, J., Sun, J., Yang, E., & Zhou, H. (2017) · Neurocomputing · doi:10.1016/j.neucom.2017.06.004 · Cited by 52
“The idea of OP is consistent with human 25 cognitive behaviors where humans scan objects before discriminating them [6].”
BackgroundHeadNet: An End-to-End Adaptive Relational Network for Head Detection
Li, W., Li, H., Wu, Q., Meng, F., Xu, L., & Ngan, K. N. (2019) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2019.2890840 · Cited by 53
“As mentioned in [67], the average recall between 0.”
BackgroundReview of Vehicle Detection Systems in Advanced Driver Assistant Systems
Sakhare, K. V., Tewari, T., & Vyas, V. (2019) · Archives of Computational Methods in Engineering · doi:10.1007/s11831-019-09321-3 · Cited by 55
“Recent times witness an OPM for object proposal generation accompanied with the classifier for object detection instead of sliding window approach to reduce the processing delay [85].”
BackgroundMethodInfluentialVehicle Detection Using Alex Net and Faster R-CNN Deep Learning Models: A Comparative Study
Espinosa, J. E., Velastín, S. A., & Branch, J. W. (2017) · Lecture notes in computer science · doi:10.1007/978-3-319-70010-6_1 · Cited by 54
Comprehensive Analysis of Deep Learning-Based Vehicle Detection in Aerial Images
Sommer, L. W., Schuchert, T., & Beyerer, J. (2018) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2018.2874396 · Cited by 50
Lan, L., Ye, C., Wang, C., & Zhou, S. (2019) · IEEE Access · doi:10.1109/access.2019.2901568 · Cited by 50
“[59], for object (WCE abnormal pattern) detection, improving region localization”
BackgroundMethodResultInfluentialDeep Learning-Based Super-Resolution Reconstruction and Marker Detection for Drone Landing
Truong, N. Q., Nguyen-Ha, P., Nam, S. H., & Park, K. R. (2019) · IEEE Access · doi:10.1109/access.2019.2915944 · Cited by 51
Multi-View Frustum Pointnet for Object Detection in Autonomous Driving
Cao, P., Chen, H., Zhang, Y., & Wang, G. (2019) · Proceedings - International Conference on Image Processing · doi:10.1109/icip.2019.8803572 · Cited by 48
“matching the two kinds of detection boxes, the intersection over union (IoU) [9] is used as criterion, if 2D IoU is above the threshold, we assume that two 2D boxes are matched.”
MethodBayesian Adaptive Superpixel Segmentation
Uziel, R., Ronen, M., & Freifeld, O. (2019) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00856 · Cited by 58
“Superpixels [37], relatively-small image segments, form a compact intermediate image representation that is a key preprocessing step in many computer-vision tasks, e.g ., [6, 14, 18, 22, 24, 25, 27, 30, 33, 34, 35, 46, 54].”
BackgroundDetecting Structural Components of Building Engineering Based on Deep-Learning Method
Hou, X., Zeng, Y., & Xue, J. (2019) · Journal of Construction Engineering and Management · doi:10.1061/(asce)co.1943-7862.0001751 · Cited by 49
3D object detection for autonomous driving: Methods, models, sensors, data, and challenges
Ghasemieh, A., & Kashef, R. F. (2022) · Transportation Engineering · doi:10.1016/j.treng.2022.100115 · Cited by 60
FACET: Fairness in Computer Vision Evaluation Benchmark
Gustafson, L., Rolland, C., Ravi, N., Duval, Q., Adcock, A., Fu, C.-Y., Hall, M. A., & Ross, C. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.01863 · Cited by 95
“We treat the remaining boxes as class-agnostic, and we compute the average recall (AR) and mAR (mean average recall) metrics proposed in [50] with all predicted boxes and measure performance across the demographic attributes.”
MethodXi, X., Xia, K., Yang, Y., Du, X., & Feng, H. (2021) · Computers and Electronics in Agriculture · doi:10.1016/j.compag.2021.106506 · Cited by 47
Simple Does It: Weakly Supervised Instance and Semantic SegmentationPreprint
Khoreva, A., Benenson, R., Hosang, J., Hein, M., & Schiele, B. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1603.07485 · Cited by 50
“The intuition is that the convnet training might benefit from trading-off lower recall (more ignore pixels) for higher precision (more pixels are correctly labelled).”
BackgroundSelf-citationLearning Sampling Distributions for Efficient Object Detection
Pang, Y., Cao, J., & Li, X. (2016) · IEEE Transactions on Cybernetics · doi:10.1109/tcyb.2015.2508603 · Cited by 47
Pair-Navi: Peer-to-Peer Indoor Navigation with Mobile Visual SLAM
Dong, E., Xu, J., Wu, C., Liu, Y., & Yang, Z. (2019) · IEEE Conference on Computer Communications (INFOCOM) · doi:10.1109/infocom.2019.8737640 · Cited by 50
“The Region-based CNN (R-CNN) approach [17] leveraged candidate object regions [39] and evaluated convolutional neural networks for each Region of Interest (RoI) for object detection.”
MethodObject detection based on saturation of visual perception
Pan, C., & Yan, W. Q. (2020) · Multimedia Tools and Applications · doi:10.1007/s11042-020-08866-x · Cited by 45
“5 binary threshold of GT, the top three models are highlighted in red, green and blue) Data Set Metric DRFI[11] HDCT[25] RBD[26] BL[18] BMS[22] MILPS[9] OUR MSRA 10000 F-Measure .”
BackgroundMethodWei, H., & Kehtarnavaz, N. (2019) · Machine Learning and Knowledge Extraction · doi:10.3390/make1030044 · Cited by 52
Singh, K. K., & Lee, Y. J. (2019) · 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/cvpr.2019.00964 · Cited by 49
“Object proposal methods aim to generate candidate object regions for an ensuing detector or segmentation model; see a great survey by [13].”
The three R’s of computer vision: Recognition, reconstruction and reorganization
Malik, J., Arbeláez, P., Carreira, J., Fragkiadaki, K., Girshick, R., Gkioxari, G., Gupta, S., Hariharan, B., Kar, A., & Tulsiani, S. (2016) · Pattern Recognition Letters · doi:10.1016/j.patrec.2016.01.019 · Cited by 54
Robust detection of non-motorized road users using deep learning on optical and LIDAR data
Kim, T., & Ghosh, J. (2016) · IEEE International Conference on Intelligent Transportation Systems (ITSC) · doi:10.1109/itsc.2016.7795566 · Cited by 49
“First, a cyclist video contains both bicycle and a human rider, and motion speed is significantly faster than the pedestrian’s walking speed.”
BackgroundHou, Y., Zhang, H., & Zhou, S. (2017) · Autonomous Robots · doi:10.1007/s10514-017-9684-3 · Cited by 43
“Although both BING and EdgeBoxes have demonstrated similar performance under different imaging conditions (e.g. illumination, rotation, scaling, etc.), BING is still slightly better than EdgeBoxes in some cases (Hosang et al. 2016).”
BackgroundTowards Dense Object Tracking in a 2D Honeybee Hive
Bożek, K., Hebert, L., Mikheyev, A. S., & Stephens, G. J. (2018) · 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/cvpr.2018.00440 · Cited by 42
Where are the Masks: Instance Segmentation with Image-level SupervisionPreprint
Laradji, I., Vázquez, D., & Schmidt, M. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1907.01430 · Cited by 53
“Many such methods also rely on object proposals [17] to ease the task of detec-”
MethodScaleNet: Guiding Object Proposal Generation in Supermarkets and Beyond
Qiao, S., Shen, W., Qiu, W., Liu, C., & Yuille, A. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.199 · Cited by 41
Instance Neural Radiance Field
Liu, Y., Hu, B., Huang, J., Tai, Y., & Tang, C. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00079 · Cited by 65
Metadata
- Authors
- Hosang, J., Benenson, R., Dollar, P., & Schiele, B.
- Affiliations
- Max Planck Institute for Informatics (DE)
- ORCID
- Schiele, Bernt
- Venue
- IEEE Transactions on Pattern Analysis and Machine Intelligence
- Published
- 2016-04-01
- Volume
- 38
- Issue
- 4
- Pages
- 814-830
- ISSN
- 0162-8828 (ISSN-L)
- 2160-9292
- OpenAlex
- W1958328135
- Cited by
- 727
- References
- 68
- Citations by year
- 2026: 9 · 2025: 27 · 2024: 34 · 2023: 36 · 2022: 57 · 2021: 58 · 2020: 63 · 2019: 95 · 2018: 93 · 2017: 126 · 2016: 98
- FWCI
- 100.45
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- IEEE Transactions on Pattern Analysis and Machine Intelligence (publishedVersion)
- PubMed (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- MPG.PuRe (Max Planck Society) (submittedVersion, other-oa)
- MPG.PuRe (Max Planck Society) (submittedVersion)
- arXiv (Cornell University)
- 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
- object detection, PASCAL dataset, ImageNet, MS COCO, DPM, R-CNN, Fast R-CNN, sliding window search
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.