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 301–400 of 764
Synthetic-to-Real Domain Adaptation for Object Instance Segmentation
Zhang, H., Tian, Y., Wang, K., He, H., & Wang, F. (2019) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn.2019.8851791 · Cited by 20
“First, we compare the proposal quality of these methods by computing the Recall-toIoU metric [30].”
MethodCell proposal network for microscopy image analysis
Akram, S. U., Kannala, J., Eklund, L., & Heikkilä, J. (2016) · Proceedings - International Conference on Image Processing · doi:10.1109/icip.2016.7532950 · Cited by 21
“In recent years, proposal generation methods have become very popular in object detection and all current top ranked methods use object proposals [1].”
MethodFace Occlusion Detection Using Deep Convolutional Neural Networks
Xia, Y., Zhang, B., & Coenen, F. (2016) · International Journal of Pattern Recognition and Artificial Intelligence · doi:10.1142/s0218001416600107 · Cited by 25
DeepProposal: Hunting Objects by Cascading Deep Convolutional LayersPreprint
Ghodrati, A., Diba, A., Pedersoli, M., Tuytelaars, T., & Gool, L. V. (2015) · arXiv (Cornell University) · doi:10.48550/arxiv.1510.04445 · Cited by 20
Object proposal on RGB-D images via elastic edge boxes
Liu, J., Ren, T., Wang, Y., Zhong, S.-H., Bei, J., & Chen, S. (2016) · Neurocomputing · doi:10.1016/j.neucom.2016.09.111 · Cited by 19
“proposal number curve [15] (Figure 6(b) and 7(b)), and the recall vs.”
InfluentialFeng, J., Li, B., Tian, L., & Dong, C. (2020) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2020.3024919 · Cited by 19
SA-RPN: A Spacial Aware Region Proposal Network for Acne Detection
Zhang, J., Zhang, L., Wang, J., Wei, X., Li, J., Jiang, X. L., & Du, D. (2023) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2023.3304727 · Cited by 20
“…of various-size acne lesions; 2) it is a standard two-stage detection methodology that usually has considerable detection performance [16], [19]; 3) it can take advantage of instance segmentation annotations to improve the detection [38]; 4) it is robust and widely utilized for comparison…”
BackgroundMethodImproved Object Detection With Iterative Localization Refinement in Convolutional Neural Networks
Cheng, K., Chen, Y., & Fang, W. (2017) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2017.2730258 · Cited by 19
“Here, we choose to combine the former with the edge box algorithm [2] simply because it can achieve better performance on the object detection problem based on the in-depth analysis in [49].”
MethodLearning Objectness from Sonar Images for Class-Independent Object Detection
Valdenegro-Toro, M. (2019) · European Conference on Mobile Robots · doi:10.1109/ecmr.2019.8870959 · Cited by 19
“Precision is not typically evaluated as unlabeled objects are considered false positives [18] which would skew any evaluation.”
Polyhedral Conic Classifiers for Computer Vision Applications and Open Set Recognition
Çevıkalp, H., & Sağlamlar, H. (2019) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2019.2934455 · Cited by 23
“We used Fast R-CNN [44] implementation with the setup given in [47] which uses Region Proposals provided by [48].”
MethodRebooting Computing and Low-Power Image Recognition Challenge
Lu, Y.-H., Kadin, A. M., Berg, A. C., Conte, T. M., DeBenedictis, E. P., Garg, R., Gingade, G., Hoang, B., Huang, Y., Li, B., et al. (2015) · International Conference on Computer Aided Design · doi:10.1109/iccad.2015.7372672 · Cited by 19
“over, recent work has demonstrated that mAP will increase slowly with the number of region proposals [18, 13].”
BackgroundUnmanned Aerial Vehicle Object Tracking by Correlation Filter with Adaptive Appearance Model
Xue, X., Li, Y., & Shen, Q. (2018) · Sensors · doi:10.3390/s18092751 · Cited by 19
“An enormous amount of research effort has gone into vision-based tasks, such as object tracking [1–3], recognition [4,5] and saliency detection [6].”
BackgroundDeep Extreme Cut: From Extreme Points to Object SegmentationPreprint
Maninis, K.-K., Caelles, S., Pont-Tuset, J., & Gool, L. V. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1711.09081 · Cited by 25
Prior knowledge auxiliary for few-shot pest detection in the wild
Wang, X., Du, J., Xie, C., Wu, S., Ma, X., Liu, K., Dong, S., & Chen, T. (2023) · Frontiers in Plant Science · doi:10.3389/fpls.2022.1033544 · Cited by 19
“AverageRecall = 2 Z 1 0:5 recall xð Þdx = 2no n i=1 max IoU gtið Þ − 0:5, 0ð Þ (6) The definition of AverageRecall (AR) is first proposed by Hosang et al. (2015), and it can be calculated using Equation (6).”
BackgroundMa, J., Ming, A., Huang, Z., Wang, X., & Zhou, Y. C. (2017) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.527 · Cited by 18
“Thirdly, the average recall (AR) between IoU 0.5 to 1 is introduced in [13] to summarize proposal performance across IoU thresholds, varying from 10 to 10000 proposals too.”
BackgroundMethodInfluentialToward Filament Segmentation Using Deep Neural Networks
Ahmadzadeh, A., Mahajan, S. S., Kempton, D., Angryk, R. A., & Ji, S. (2019) · IEEE International Conference on Big Data (BigData) · doi:10.1109/bigdata47090.2019.9006340 · Cited by 18
“This is similar to what was proposed in [17], except that in COCO it is averaged over all categories.”
MethodInsulator Defect Detection Based on Improved Faster R-CNN
Tang, J., Wang, J., Wang, H., Wei, J., Wei, Y., & Qin, M. (2022) · Asia Energy and Electrical Engineering Symposium (AEEES) · doi:10.1109/aeees54426.2022.9759683 · Cited by 18
Design Features of Grocery Product Recognition Using Deep Learning
Gothai, E., Bhatia, S., Alabdali, A. M., Sharma, D. K., Kondamudi, B. R., & Dadheech, P. (2022) · Intelligent Automation & Soft Computing · doi:10.32604/iasc.2022.026264 · Cited by 18
Egocentric Temporal Action Proposals
Huang, S., Wang, W., He, S., & Lau, R. W. H. (2017) · IEEE Transactions on Image Processing · doi:10.1109/tip.2017.2772904 · Cited by 20
“A comprehensive survey can be found in [20].”
BackgroundLearning to Segment Object Candidates via Recursive Neural Networks
Chen, T., Lin, L., Wu, X., Xiao, N., & Luo, X. (2018) · IEEE Transactions on Image Processing · doi:10.1109/tip.2018.2859025 · Cited by 18
“This is widely adopted by many proposal works [26], [40] for”
InfluentialPig Detection Algorithm Based on Sliding Windows and PCA Convolution
Sun, L., Liu, Y., Chen, S., Luo, B., Li, Y., & Liu, C. (2019) · IEEE Access · doi:10.1109/access.2019.2907748 · Cited by 18
BING++: A Fast High Quality Object Proposal Generator at 100fpsPreprint
Zhang, Z., Liu, Y., Bolukbasi, T., Cheng, M., & Saligrama, V. (2015) · arXiv (Cornell University) · Cited by 18
Detecting temporally consistent objects in videos through object class label propagation
Tripathi, S., Belongie, S., Hwang, Y., & Nguyen, T. Q. (2016) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv.2016.7477702 · Cited by 20
“A comparative literature survey on object proposal methods and their evaluations can be found in [15, 14].”
BackgroundDetection and tracking based tubelet generation for video object detection
Wang, B., Tang, S., Xiao, J.-B., Quan-feng, Y., & Zhang, Y. (2018) · Journal of Visual Communication and Image Representation · doi:10.1016/j.jvcir.2018.11.014 · Cited by 17
Robust and efficient multi-source image matching method based on best-buddies similarity measure
Wu, Q., Xu, G., Cheng, Y., Wang, Z., Dong, W., & Ma, L. (2019) · Infrared Physics & Technology · doi:10.1016/j.infrared.2019.05.020 · Cited by 17
MU R-CNN: A Two-Dimensional Code Instance Segmentation Network Based on Deep Learning
Yuan, B., Li, Y., Jiang, F., Xu, X., Guo, Y., Zhao, J., Zhang, D., Guo, J., & Shen, X. (2019) · Future Internet · doi:10.3390/fi11090197 · Cited by 17
“According to Hosang [40], methods such as Selective Search and Edge Boxes seem to strike a better balance between recall and repeatability, if precise proposals are required.”
Enhancing semantic segmentation with detection priors and iterated graph cuts for robotics
Antonello, M., Chiesurin, S., & Ghidoni, S. (2020) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2019.103467 · Cited by 18
Object detection through region proposal based techniques
Prakash, S. R., & Singh, P. N. (2021) · Materials Today Proceedings · doi:10.1016/j.matpr.2021.02.533 · Cited by 17
Robust Ship Detection in Infrared Images through Multiscale Feature Extraction and Lightweight CNN
Miao, R., Jiang, H., & Tian, F. (2022) · Sensors · doi:10.3390/s22031226 · Cited by 17
“For example, some researchers propose a saliency strategy, a feature descriptor [7,8], and a local comparison method [9] to determine small infrared targets, but these methods are very sensitive to noise, which usually generates a high false-alarm rate.”
BackgroundDynamic Multitarget Detection Algorithm of Voxel Point Cloud Fusion Based on PointRCNN
Luo, X., Zhou, F., Tao, C., Yang, A., Zhang, P., & Chen, Y. (2022) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2022.3176390 · Cited by 18
“Hosang et al. [3] analyzed the difficulties of target detection technology, and Chen et al. [4] analyzed mainstream architectures of target detection and typical feature extraction networks.”
BackgroundSaif, A. F. M. S., & Mahayuddin, Z. R. (2022) · International Journal of Advanced Computer Science and Applications · doi:10.14569/ijacsa.2022.0131123 · Cited by 23
“1) Oracle Recall: Oracle recall computes the percentage of the recalled ground truth objects to receive recall rate [75, 2, 76, 104,105].”
BackgroundSemi-supervised object detection based on single-stage detector for thighbone fracture localization
Wei, J., Yao, J., Zhang, G., Guan, B., Zhang, Y., & Wang, S. (2023) · Neural Computing and Applications · doi:10.1007/s00521-023-09277-3 · Cited by 19
“AP (average precision) is the mainstream evaluation index of object detection model [54].”
Gillert, A., Resente, G., Anadon‐Rosell, A., Wilmking, M., & von Lukas, U. F. (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/cvpr52729.2023.01397 · Cited by 16
Detect2Rank: Combining Object Detectors Using Learning to Rank
Karaoğlu, S., Liu, Y., & Gevers, T. (2015) · IEEE Transactions on Image Processing · doi:10.1109/tip.2015.2499702 · Cited by 18
Fisher-selective search for object detection
Buzcu, I., & Alatan, A. A. (2016) · Proceedings - International Conference on Image Processing · doi:10.1109/icip.2016.7533037 · Cited by 16
“Per [9], these can be divided into two groups: window-scoring and grouping based methods.”
BackgroundGeneric Proposal Evaluator: A Lazy Learning Strategy Toward Blind Proposal Quality Assessment
Wu, Q., Li, H., Meng, F., & Ngan, K. N. (2017) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2017.2750070 · Cited by 18
“Unsurprisingly, the proposed GPE still achieves the highest ABO result.”
Learn to Detect Objects Incrementally
Guan, L., Wu, Y. X., Zhao, J., & Ye, C. (2018) · IEEE Intelligent Vehicles Symposium · doi:10.1109/ivs.2018.8500673 · Cited by 19
“But category-agnostic proposals are inefficient [29], so this method cannot be used in scenes requiring real-time detect speed, such as autopilot environment.”
MethodChai, D. (2020) · International Journal of Computer Vision · doi:10.1007/s11263-020-01352-9 · Cited by 16
“…2016), semantic segmentation (Farabet et al. 2012; Mostajabi et al. 2015; Gadde et al. 2016), stereo computation (Miˇcušík and Košecká 2010; Guney and Geiger 2015), object tracking (Wang et al. 2011), objectness measuring (Alexe et al. 2012), object proposal generation (Hosang et al. 2015), etc.”
BackgroundBoundary-aware box refinement for object proposal generation
Chen, X., Ma, H., Zhu, C., Wang, X., & Zhao, Z. (2016) · Neurocomputing · doi:10.1016/j.neucom.2016.09.045 · Cited by 15
Large field-of-view pine wilt disease tree detection based on improved YOLO v4 model with UAV images
Zhang, Z., Han, C., Wang, X., Li, H., Li, J., Zeng, J., Sun, S., & Wu, W. (2024) · Frontiers in Plant Science · doi:10.3389/fpls.2024.1381367 · Cited by 15
“The focus of the research is how to change the structure of the algorithm model and achieve a balance between detection accuracy and processing time (Hosang et al., 2016).”
BackgroundCNN-based edge filtering for object proposals
Waris, M. A., Iosifidis, A., & Gabbouj, M. (2017) · Neurocomputing · doi:10.1016/j.neucom.2017.05.071 · Cited by 15
“Following [23, 24], we also evaluate the quality of object proposals using the recall metric.”
BackgroundMethodRefinedbox: Refining for fewer and high-quality object proposals
Liu, Y., Li, S., & Cheng, M. (2020) · Neurocomputing · doi:10.1016/j.neucom.2020.04.017 · Cited by 15
“[37] to calculate the average recall when IoU thresholds are [0.”
BackgroundTaghizadeh, M., & Chalechale, A. (2021) · Expert Systems with Applications · doi:10.1016/j.eswa.2021.116105 · Cited by 16
Tao, C., Shi-ping, F., Wang, C., Luo, X., Li, H., Gao, Z., Zhang, Z., & Zheng, S. (2022) · IEEE Internet of Things Journal · doi:10.1109/jiot.2022.3231369 · Cited by 15
“D OBJECT detection is one of the essential functions for autonomous vehicles to sense a surrounding environment, which has aroused extensive attention in recent years [1], [2], [3], [4], [5], [6], [7].”
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
“Since initial box quality is critical to WSOD performance [31], using a heuristic to generate proposals is often insufficient and impractical.”
BackgroundGrard, M., Brégier, R., Sella, F., Dellandréa, E., & Chen, L. (2018) · Springer proceedings in advanced robotics · doi:10.1007/978-3-319-89327-3_16 · Cited by 15
Saliency-guided adaptive seeding for supervoxel segmentation
Gao, G., Lauri, M., Zhang, J., & Frintrop, S. (2017) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2017.8206374 · Cited by 15
“A popular application of superpixels is object discovery, where superpixels are grouped to obtain object candidates which in turn are used as input for object recognition methods [14], [9], [10].”
Background(2015) · arXiv (Cornell University) · Cited by 14
“These methods, which are extensively reviewed in [13], originate in the idea of “objectness” proposed by [1].”
BackgroundMethodActive Object Detection Using Double DQN and Prioritized Experience Replay
Han, X., Liu, H., Sun, F., & Yang, D. (2018) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn.2018.8489296 · Cited by 14
“In object detection, a bounding box is a region proposal, there are many works focusing on this problem [20].”
BackgroundLearning to Co-Generate Object Proposals with a Deep Structured Network
Hayder, Z., He, X., & Salzmann, M. (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.2016.281 · Cited by 14
Robust Double-Encoder Network for RGB-D Panoptic Segmentation
Sodano, M., Magistri, F., Guadagnino, T., Behley, J., & Stachniss, C. (2023) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra48891.2023.10160315 · Cited by 16
“Their goal is to extract a number of candidate object regions [12] [18], and then evaluate them independently.”
BackgroundYarn Density Measurement for 3-D Braided Composite Preforms Based on Rotation Object Detection
Dai, L., Li, D., Tu, Q., & Wang, J. (2022) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2022.3192294 · Cited by 14
Segmentation Free Object Discovery in Video
Cuffaro, G., Becattini, F., Baecchi, C., Seidenari, L., & Bimbo, A. D. (2016) · Lecture notes in computer science · doi:10.1007/978-3-319-49409-8_4 · Cited by 13
“Object proposals [3] provide a relatively small set of bounding boxes likely to contain salient regions in images, based on some objectness measure.”
BackgroundChen, C., Shen, F., & Dai, C. (2023) · Measurement · doi:10.1016/j.measurement.2023.112953 · Cited by 13
Yang, J., & Gapar, Y. (2024) · ICCK Transactions on Emerging Topics in Artificial Intelligence · doi:10.62762/tetai.2024.115892 · Cited by 13
“Following literature [15], prediction callbacks are used as a performance metric.”
Proposal generation method for object detection in infrared image
Sun, N., Jiang, F., Yan, H.-C., Liu, J., & Han, G. (2016) · Infrared Physics & Technology · doi:10.1016/j.infrared.2016.12.021 · Cited by 13
A Pedestrian Detection System Accelerated by Kernelized Proposals
Baek, J., Hyun, J., & Kim, E. (2019) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2019.2904836 · Cited by 13
“These grouping methods demonstrate high recall performance for object proposals; however, they are unsuitable for real-time PD because their computation time is usually greater than 1 s per frame [10].”
BackgroundMethodInfluentialA New Region Proposal Network for Far-Infrared Pedestrian Detection
Cao, Z., Yang, H., Zhao, J., Pan, X., Zhang, L., & Liu, Z. (2019) · IEEE Access · doi:10.1109/access.2019.2932749 · Cited by 13
Detecting plains and Grevy's Zebras in the realworld
Parham, J. R., & Stewart, C. V. (2016) · IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW) · doi:10.1109/wacvw.2016.7470122 · Cited by 13
Hou, Y., Zhang, H., Zhou, S., & Zou, H. (2017) · Mobile Information Systems · doi:10.1155/2017/8104386 · Cited by 13
“We prefer BING for the following three reasons: (a) as has been demonstrated in [42], compared to EdgeBoxes, BING has slightly better repeatability, which is an important property for localization accuracy; (b) our previous experimental evaluation also shows that the localization accuracy achieved by BING is comparabl…”
ResultUlmamei, A., D’Adamo, T., Vasile, C.-E., & Hobincu, R. (2025) · Journal of Imaging · doi:10.3390/jimaging11120436 · Cited by 12
“AR, introduced by Hosang et al. [60], is also widely used in COCO challenges.”
Context Proposals for Saliency Detection
Azaza, A., van de Weijer, J., Douik, A., & Masana, M. (2018) · Computer Vision and Image Understanding · doi:10.1016/j.cviu.2018.06.002 · Cited by 12
“They propose either a set of bounding boxes or image segments, which have a high probability of containing an object [19, 27].”
BackgroundMethodMethod for Vortex Shape Retrieval and Area Calculation Based on Convex Hull Algorithm
Wei, X., Li, J., Long, B., Hu, X., Wu, H., & Li, H. (2020) · IEEE Access · doi:10.1109/access.2020.3046651 · Cited by 12
“suggested an object selection algorithm using binarized normed gradient to screen candidate areas based on the edge detector [8].”
MethodAutomatic MEP Component Detection with Deep Learning
Kufuor, J., Mohanty, D. D., Valero, E., & Bosché, F. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-68787-8_28 · Cited by 12
Finding Deviated Behaviors of the Compressed DNN Models for Image Classifications
Tian, Y., Zhang, W., Wen, M., Cheung, S.-C., Sun, C. P., Ma, S., & Jiang, Y. (2023) · ACM Transactions on Software Engineering and Methodology · doi:10.1145/3583564 · Cited by 16
“For example, in object detection, the location and boundary of the detected object are also important [27].”
Bouma, H., de Boer, M., Kruithof, M. C., ter Haar, F. B., Fischer, N. M., Hagendoorn, L., Joosten, B., & Raaijmakers, S. A. (2017) · Counterterrorism, Crime Fighting, Forensics, and Surveillance Technologies · doi:10.1117/12.2277970 · Cited by 14
Una Revisión Sistemática de Métodos para Localizar Automáticamente Objetos en Imágenes
Chaves, D., Saikia, S., Fernández-Robles, L., Alegre, E., & Trujillo, M. (2018) · Revista Iberoamericana de Automática e Informática Industrial RIAI · doi:10.4995/riai.2018.10229 · Cited by 15
“Generalmente, se utilizan clasificadores más complejos, en comparación con los que se usan con ventanas deslizantes, mejorando los tiempos de ejecución sin afectar la precisión en la detección (Hosang et al., 2016).”
BackgroundFine-grained vehicle recognition by deep Convolutional Neural Network
Huang, K., & Zhang, B. (2016) · International Congress on Image and Signal Processing (CISP) · doi:10.1109/cisp-bmei.2016.7852756 · Cited by 16
“Image features are extracted from object proposals algorithms like selective search [23,26], and then classified with class specific linear SVMs.”
MethodSmall Object Segmentation Using Dilated Convolutions With Increasing-Decreasing Dilation
Hamaguchi, R., Fujita, A., & Nemoto, K. (2024) · IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing · doi:10.1109/jstars.2024.3477606 · Cited by 11
LAVID: A Lightweight and Autonomous Smart Camera System for Urban Violence Detection and Geolocation
Azzakhnini, M., Saidi, H., Azough, A., Tairi, H., & Qjidaa, H. (2025) · Computers · doi:10.3390/computers14040140 · Cited by 11
Elastic Edge Boxes for Object Proposal on RGB-D Images
Liu, J., Ren, T., & Bei, J. (2016) · Lecture notes in computer science · doi:10.1007/978-3-319-27671-7_17 · Cited by 11
“As shown in [9], improving intersection over union (IoU) of candidate regions and objects is as important as increasing recall of objects in several applications, such as object detection.”
BackgroundMethodInfluentialAutomatic Human Detection Using Reinforced Faster-RCNN for Electricity Conservation System
Ushasukhanya, S., & Karthikeyan, M. (2021) · Intelligent Automation & Soft Computing · doi:10.32604/iasc.2022.022654 · Cited by 11
“Secondly, Fast RCNN makes use of a bounding box regressor to optimize the region proposals, which yields high accuracy and also speeds up the training process [12].”
MethodAn Real Time Object Detection Method for Visually Impaired Using Machine Learning
Alagarsamy, S., Rajkumar, T. D., Syamala, K. P. L., Niharika, C. S., Rani, D. U., & Balaji, K. (2023) · International Conference on Computer Communication and Informatics (ICCCI) · doi:10.1109/iccci56745.2023.10128388 · Cited by 11
“Humans require vision as a sense since it is essential to how they interpret their surroundings [2].”
BackgroundZhang, X., Zhang, J., Chen, J., Guo, R., & Wu, J. (2023) · The Visual Computer · doi:10.1007/s00371-023-02980-1 · Cited by 12
“Increasing the number of proposals can result in a promising recall, but this will significantly lengthen the detection’s running time [40].”
Karianakis, N., Dong, J., & Soatto, S. (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.2016.481 · Cited by 13
“Such intense interest was sparked by the surprising performance of CNNs [6, 11, 19, 23, 28, 36, 37, 40, 42] in Computer Vision benchmarks [10, 15], where many couple a proposal scheme [1, 5, 7, 14, 24, 25, 27, 31, 35, 44, 48] with a CNN.”
BackgroundMethodInfluentialEfficient object proposals extraction for target detection in VHR remote sensing images
Farooq, A., Hu, J., & Jia, X. (2017) · IEEE International Geoscience and Remote Sensing Symposium (IGARSS) · doi:10.1109/igarss.2017.8127712 · Cited by 11
“Using the sliding window, Gradient-based methods have been successfully implemented by the researchers to detect the vehicles [5].”
BackgroundMethodDeep multiple instance learning for airplane detection in high-resolution imagery
Mohammadi, M. R. (2021) · Machine Vision and Applications · doi:10.1007/s00138-020-01153-7 · Cited by 11
Multi-peak Graph-based Multi-instance Learning for Weakly Supervised Object Detection
Ji, R., Liu, Z., Zhang, L., Liu, J., Zuo, X., Wu, Y., Zhao, C., Wang, H., & Yang, L. (2021) · ACM Transactions on Multimedia Computing Communications and Applications · doi:10.1145/3432861 · Cited by 11
“References [7, 31, 32] make detailed comparisons and discussions about object proposal methods.”
BackgroundAn efficient deep neural network with color-weighted loss for fire detection
Zhang, R., Zhang, W., Liu, Y., Li, P., & Zhao, J. (2022) · Multimedia Tools and Applications · doi:10.1007/s11042-022-12861-9 · Cited by 11
A Multi-strategy Region Proposal Network
Chen, Y., Li, Y., Wang, G., & Xu, Q. (2018) · Expert Systems with Applications · doi:10.1016/j.eswa.2018.06.043 · Cited by 11
Fast template matching based on deformable best-buddies similarity measure
Xia, H., Zhao, W., Jiang, F., Li, H.-S., Xin, J. L., & Zhou, Z. (2018) · Multimedia Tools and Applications · doi:10.1007/s11042-018-6722-x · Cited by 11
Image-Based Goat Breed Identification and Localization Using Deep Learning
Ghosh, P., Mustafi, S., & Mandal, S. N. (2020) · International Journal of Computer Vision and Image Processing · doi:10.4018/ijcvip.2020100105 · Cited by 11
Large Scale Asset Extraction for Urban Images
Affara, L., Nan, L., Ghanem, B., & Wonka, P. (2016) · Lecture notes in computer science · doi:10.1007/978-3-319-46487-9_27 · Cited by 11
“For a comprehensive evaluation and an in-depth analysis of object proposal methods, please refer to the recent survey [11].”
MethodA Two-Stage Insulator Defect Detection Network With Sequence Transduction
Zhang, Y. H., Wang, B., Yang, Q., Tang, F., & Wei, K. (2025) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2024.3522390 · Cited by 15
“2) Evaluation Metric: To measure the effectiveness of our proposed model, the widely used AP, recall, and AR [27] are adopted, which are expressed by where TP , FP , FN , P , R represent true positive, false positive, false negative, precision, and recall, respectively.”
Ping, P., Zhang, X., Lu, T., Shi, Q., Tian, Y., Yan, J., & Ding, W. (2025) · Information Fusion · doi:10.1016/j.inffus.2025.103653 · Cited by 20
Deep learning object detection applied to defect recognition of memory modules
Huang, J.-T., & Ting, C.-H. (2022) · The International Journal of Advanced Manufacturing Technology · doi:10.1007/s00170-022-09716-w · Cited by 10
“…reduce the memory module defects caused by handling the assembly as well as the imperfect quality inspection of memory modules purchased from suppliers, AI technology can be effectively used in actual products [1], which has always been the goal of the server industry’s continuous efforts.”
BackgroundDeep Free-Form Deformation Network for Object-Mask Registration
Zhang, H., & He, X. (2017) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.456 · Cited by 10
“The AR metric describes the overall quality of object proposals and has been shown to correlate highly with the detection accuracy in [15].”
BackgroundMethodA Deep Learning-Based System for Monitoring the Number and Height Growth Rates of Moso Bamboo Shoots
Hong, S., Jiang, Z., Zhu, J., Rao, Y., Wu, Z., & Gao, J. (2022) · Applied Sciences · doi:10.3390/app12157389 · Cited by 10
Automatic wildfire monitoring system based on deep learning
Peng, Y., & Wang, Y. (2022) · European Journal of Remote Sensing · doi:10.1080/22797254.2022.2133745 · Cited by 10
“Even if the image size is increased, the speed of the algorithm increases linearly at most which is acceptable (Hosang et al., 2016).”
BackgroundLiu, C., Ding, Y., Zhu, M. B., Xiu, J., Li, M., & Li, Q. (2019) · Sensors · doi:10.3390/s19153294 · Cited by 10
“5 to 1, a metric namely average recall (AR) [23] was introduced, which calculates recalls under different IoU thresholds and performs with an average value.”
BackgroundMethodInfluentialN-RPN: Hard Example Learning For Region Proposal Networks
Cho, M., Chung, T.-Y., Lee, H., & Lee, S. (2019) · Proceedings - International Conference on Image Processing · doi:10.1109/icip.2019.8803519 · Cited by 10
“For comparing models, we use Bing [13], CPMC [14], EdgeBoxes [2], Endres [15], Geodesic [16], MCG [17], Objectness [18], Rahtu [19], RandomizedPrims [5], Rantalankila [20], Rigor [21], SelectiveSearch [1], Gaussian, SlidingWindow, Superpixels, Uiform [5], RPN [3].”
MethodHierarchical Region Proposal Refinement Network for Weakly Supervised Object Detection
Zhang, M., Liu, S., & Zeng, B. (2021) · Proceedings - International Conference on Image Processing · doi:10.1109/icip42928.2021.9506087 · Cited by 10
“However, the quality of proposals can influence the training of the detector and limit the detection performance [15, 9].”
BackgroundRoad Object Detection Using a Disparity-Based Fusion Model
Chen, J., Xu, W., Peng, W., Bu, W., Xing, B., & Liu, G. (2018) · IEEE Access · doi:10.1109/access.2018.2825229 · Cited by 10
“same [28].”
YOLO-lychee-advanced: an optimized detection model for lychee pest damage based on YOLOv11
Wu, X., Su, X., Ma, Z., & Xu, B. (2025) · Frontiers in Plant Science · doi:10.3389/fpls.2025.1643700 · Cited by 9
“…the proposed model, our model adopts precision (P), recall (R), mean average precision (mAP), and mAP50 – 95 as the primary evaluation et al., 2010; Hosang et al., 2016), de fi ned as follows: Precision (P): Re fl ects the proportion of samples predicted as positive that are truly positive,…”
Deep understanding of big geospatial data for self-driving: Data, technologies, and systems
Wang, H., Feng, J., Li, K., & Chen, L. (2022) · Future Generation Computer Systems · doi:10.1016/j.future.2022.07.003 · Cited by 9
3-D Dynamic Multitarget Detection Algorithm Based on Cross-View Feature Fusion
Zhou, F., Tao, C., Gao, Z., Zhang, Z., Zheng, S., & Zhu, Y. (2023) · IEEE Transactions on Artificial Intelligence · doi:10.1109/tai.2023.3342104 · Cited by 10
“Moreover, the detection performance has received extensive attention in recent years [8], [9].”
BackgroundRelief R-CNN: Utilizing Convolutional Features for Fast Object Detection
Li, G., Liu, J., Jiang, C., Zhang, L., Lin, M., & Tang, K. (2017) · Lecture notes in computer science · doi:10.1007/978-3-319-59072-1_46 · Cited by 9
“comes from three parts: 1) the iterative RoIs generation process [12]; 2) the deep CNN with a huge computation requirement [15,22,10]; and 3) the naive combination of RoIs and the deep CNN [8].”
InfluentialZhang, M., Yang, Y., Zhang, H., Ji, Y., Xie, N., & Shen, H. T. (2017) · Lecture notes in computer science · doi:10.1007/978-3-319-68155-9_20 · Cited by 9
Toward Scale-Invariance and Position-Sensitive Region Proposal Networks
Lu, H.-F., Du, X., & Chang, P.-L. (2018) · Lecture notes in computer science · doi:10.1007/978-3-030-01237-3_11 · Cited by 9
An Enhanced Histogram of Oriented Gradient Descriptor for Numismatic Applications
Hmood, A. K., Suen, C. Y., & Lam, L. (2018) · Pattern Recognition and Image Analysis · doi:10.1134/s1054661818040028 · Cited by 9
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.