Cited by 53,803Advanced Neural Network ApplicationsAdvanced Image and Video Retrieval TechniquesVisual Attention and Saliency Detection

You Only Look Once: Unified, Real-Time Object Detection

Open access
Redmon, Joseph, Divvala, Santosh, Girshick, Ross, Farhadi, Ali·2016·2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

Abstract

We present YOLO, a new approach to object detection. Prior work on object detection repurposes classifiers to perform detection. Instead, we frame object detection as a regression problem to spatially separated bounding boxes and associated class probabilities. A single neural network predicts bounding boxes and class probabilities directly from full images in one evaluation. Since the whole detection pipeline is a single network, it can be optimized end-to-end directly on detection performance. Our unified architecture is extremely fast. Our base YOLO model processes images in real-time at 45 frames per second. A smaller version of the network, Fast YOLO, processes an astounding 155 frames per second while still achieving double the mAP of other real-time detectors. Compared to state-of-the-art detection systems, YOLO makes more localization errors but is less likely to predict false positives on background. Finally, YOLO learns very general representations of objects. It outperforms other detection methods, including DPM and R-CNN, when generalizing from natural images to other domains like artwork.

AI snapshot

Written from the paper’s full text.

TL;DR

YOLO reframes object detection as a single regression problem, using one convolutional network to predict bounding boxes and class probabilities directly from full images. It achieves real-time detection speeds while remaining competitive with, and complementary to, slower region-based detectors.

Key findings

  • YOLO runs at 45 FPS on a Titan X GPU, and Fast YOLO runs at 155 FPS; Fast YOLO reaches 52.7% mAP on PASCAL VOC 2007, more than double prior real-time detectors, while YOLO reaches 63.4% mAP.
  • On PASCAL VOC 2012, YOLO scores 57.9% mAP; combining Fast R-CNN with YOLO raises Fast R-CNN from 71.8% to 75.0% mAP on VOC 2007, a 3.2% gain.
  • YOLO makes fewer background false positives than Fast R-CNN: Fast R-CNN has 13.6% background false positives among top detections and is almost 3× more likely to predict background detections than YOLO.
  • YOLO generalizes better to artwork: on Picasso, YOLO gets 53.3 AP versus R-CNN’s 10.4 AP and DPM’s 37.8 AP; on People-Art, YOLO gets 45 AP versus R-CNN’s 26 AP and DPM’s 32 AP.

Methods

  • Detection is modeled as a single CNN regression: the image is divided into an S×S grid; for PASCAL VOC, S=7, B=2, C=20, producing a 7×7×30 tensor of bounding boxes, confidence scores, and class probabilities.
  • The base network has 24 convolutional layers followed by 2 fully connected layers; Fast YOLO uses 9 convolutional layers. Convolutional layers are pretrained on ImageNet at 224×224, then detection training uses 448×448 input.
  • Training uses PASCAL VOC 2007 and 2012 for about 135 epochs, batch size 64, momentum 0.9, decay 0.0005, sum-squared error with λ_coord=5 and λ_noobj=0.5, dropout, and data augmentation.

Limitations

  • The paper states YOLO imposes strong spatial constraints because each grid cell predicts only two boxes and one class, limiting nearby objects and struggling with small objects in groups such as flocks of birds.
  • The paper states YOLO struggles to generalize to objects in new or unusual aspect ratios or configurations, uses relatively coarse features due to downsampling, and its loss treats small and large box errors equally; its main error source is incorrect localization.
  • The paper states YOLO still lags state-of-the-art detection accuracy and struggles to precisely localize some objects, especially small ones.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1109/cvpr.2016.91

Cites

52 of 52 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

7 of 25 cited sources checked · 2 supported · 0 flagged · accuracy not yet measured

  • Premisecitation
    § 4.1. Comparison to Other Real-Time Systems · p. 5

    R-CNN minus R falls short of real-time.

    “While it is much faster than R-CNN, it still falls short of real-time and takes a significant accuracy hit from not having good proposals.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [20] [20] K. Lenc and A. Vedaldi. R-cnn minus r. arXiv preprint arXiv:1506.06981 , 2015. 5, 6 - source

      The ConvNet-based Object Detection section identifies R-CNN minus R as a method that uses predefined regions instead of region proposals. It gives no runtime measurement or assessment of whether the method achieves real-time detection; searches for real-time performance and speed did not establish that claim. (LLM, conf=0.76)

      Experimental — accuracy not yet measured
  • Study design and conductcitation
    § 2.2. Training · p. 4

    To avoid overfitting, YOLO uses dropout and extensive data augmentation.

    “To avoid overfitting we use dropout and extensive data augmentation.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [18] [18] G. E. Hinton, N. Srivastava, A. Krizhevsky, I. Sutskever, and R. R. Salakhutdinov. Improving neural networks by preventing co-adaptation of feature detectors. arXiv preprint arXiv:1207.0580 , 2012. 4 - source

      The paper states that dropout reduces overfitting (p. 1) and describes using dropout with data augmentation for its ImageNet model (H Models for ImageNet, p. 18). It does not establish that YOLO uses either technique or describe YOLO’s augmentation as extensive. (LLM, conf=0.88)

      Experimental — accuracy not yet measured
  • Premisecitation
    § 4.1. Comparison to Other Real-Time Systems · p. 5

    R-CNN minus R is much faster than R-CNN.

    “While it is much faster than R-CNN, it still falls short of real-time and takes a significant accuracy hit from not having good proposals.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [20] [20] K. Lenc and A. Vedaldi. R-cnn minus r. arXiv preprint arXiv:1506.06981 , 2015. 5, 6 - source

      The ConvNet-based Object Detection section identifies R-CNN minus R and says it uses predefined regions in an SPPnet system, but gives no speed comparison with R-CNN. A search for R-CNN minus R and speed found no passage in our copy establishing that it is much faster. (LLM, conf=0.73)

      Experimental — accuracy not yet measured
  • Premisecitation
    § 4.1. Comparison to Other Real-Time Systems · p. 5

    R-CNN minus R takes a significant accuracy hit from not having good proposals.

    “While it is much faster than R-CNN, it still falls short of real-time and takes a significant accuracy hit from not having good proposals.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [20] [20] K. Lenc and A. Vedaldi. R-cnn minus r. arXiv preprint arXiv:1506.06981 , 2015. 5, 6 - source

      The “ConvNet-based Object Detection” section (page unavailable) says R-CNN minus R uses predefined regions instead of region proposals. It does not say that this causes a significant accuracy loss; the paper’s discussion of low accuracy concerns a different system, naïve Faster R-CNN. (LLM, conf=0.83)

      Experimental — accuracy not yet measured
  • Findingown data
    § 4.4. VOC 2012 Results · p. 7

    YOLO struggles with small objects compared to its closest competitors.

    “Our system struggles with small objects compared to its closest competitors.”
    Quote in the paperNot a substantial claim
  • Classified as probably not a substantial claim. 17
  • May combine two claims. 15
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 2.2. Training · p. 3

    Adding both convolutional and connected layers to pretrained networks can improve performance.

    “Ren et al. show that adding both convolutional and connected layers to pretrained networks can improve performance [29].”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [29] [29] S. Ren, K. He, R. B. Girshick, X. Zhang, and J. Sun. Object detection networks on convolutional feature maps. CoRR , abs/1504.06066, 2015. 3, 7 - source
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 4.1. Comparison to Other Real-Time Systems · p. 5

    R-CNN minus R replaces Selective Search with static bounding box proposals.

    “R-CNN minus R replaces Selective Search with static bounding box proposals [20].”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [20] [20] K. Lenc and A. Vedaldi. R-cnn minus r. arXiv preprint arXiv:1506.06981 , 2015. 5, 6 - source
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 2.2. Training · p. 4

    A dropout layer with rate .5 after the first connected layer prevents co-adaptation between layers.

    “A dropout layer with rate = .5 after the first connected layer prevents co-adaptation between layers [18].”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [18] [18] G. E. Hinton, N. Srivastava, A. Krizhevsky, I. Sutskever, and R. R. Salakhutdinov. Improving neural networks by preventing co-adaptation of feature detectors. arXiv preprint arXiv:1207.0580 , 2012. 4 - source

      The main text (p. 1) says dropout omits hidden units with probability 0.5 to prevent complex co-adaptations. But the discussion of fully connected layers (p. 5) does not establish that placing one dropout layer after the first fully connected layer has that effect; it says dropout in all hidden layers works better than dropout in only one. (LLM, conf=0.87)

      Experimental — accuracy not yet measured
  • May combine two claims.

    Findingown data
    § 4.4. VOC 2012 Results · p. 7

    YOLO's VOC 2012 mAP is lower than the current state of the art and closer to the original R-CNN using VGG-16.

    “This is lower than the current state of the art, closer to the original R-CNN using VGG-16, see Table 3.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given
    § 1. Introduction · p. 2

    YOLO still lags behind state-of-the-art detection systems in accuracy.

    “YOLOstill lags behind state-of-the-art detection systems in accuracy.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given
    § 2.3. Inference · p. 4

    Non-maximal suppression adds 23% in mAP.

    “While not critical to performance as it is for R-CNN or DPM, non-maximal suppression adds 23% in mAP.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S4 (based).

    Findingnone given
    § 6. Conclusion · p. 8

    Fast YOLO is the fastest general-purpose object detector in the literature.

    “Fast YOLO is the fastest general-purpose object detector in the literature and YOLO pushes the state-of-the-art in real-time object detection.”
    Quote in the paperNot a substantial claim
  • May combine two claims.

    Findingreasoning
    § 3. Comparison to Other Detection Systems · p. 4

    YOLO's unified architecture leads to a faster, more accurate model than DPM.

    “Our unified architecture leads to a faster, more accurate model than DPM.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingreasoning
    § 5. Real-Time Detection In The Wild · p. 8

    When attached to a webcam, YOLO functions like a tracking system, detecting objects as they move around and change in appearance.

    “While YOLO processes images individually, when attached to a webcam it functions like a tracking system, detecting objects as they move around and change in appearance.”
    Quote in the paperNot a substantial claim
  • May combine two claims.

    Findingown data
    § 4.1. Comparison to Other Real-Time Systems · p. 5

    YOLO trained using VGG-16 is more accurate but significantly slower than YOLO.

    “We also train YOLO using VGG-16. This model is more accurate but also significantly slower than YOLO.”
    Quote in the paperNot a substantial claim
  • May combine two claims.

    Findingown data
    § 4.1. Comparison to Other Real-Time Systems · p. 5

    YOLO achieves 63.4% mAP while maintaining real-time performance.

    “YOLO pushes mAP to 63 . 4% while still maintaining real-time performance.”
    Quote in the paperSubstantial claim
  • May combine two claims.

    Findingown data
    § 4.1. Comparison to Other Real-Time Systems · p. 5

    Fast YOLO is the fastest object detection method on PASCAL; as far as the authors know, it is the fastest extant object detector.

    “Fast YOLO is the fastest object detection method on PASCAL; as far as we know, it is the fastest extant object detector.”
    Quote in the paperSubstantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given
    § 4.5. Generalizability: Person Detection in Artwork · p. 7

    R-CNN has high AP on VOC 2007.

    “R-CNN has high AP on VOC 2007.”
    Quote in the paperNot a substantial claim
  • May combine two claims.

    Findingown data
    § 5. Real-Time Detection In The Wild · p. 7

    YOLO is a fast, accurate object detector.

    “YOLO is a fast, accurate object detector, making it ideal for computer vision applications.”
    Quote in the paperSubstantial claim
  • Probably not a substantial claim: may not pass S4 (based).

    Findingnone given
    § 6. Conclusion · p. 8

    YOLO generalizes well to new domains.

    “YOLO also generalizes well to new domains making it ideal for applications that rely on fast, robust object detection.”
    Quote in the paperNot a substantial claim
  • May combine two claims.

    Interpretationreasoning
    § 2. Unified Detection · p. 2

    The YOLO design enables end-to-end training and real-time speeds while maintaining high average precision.

    “The YOLO design enables end-to-end training and realtime speeds while maintaining high average precision.”
    Quote in the paperSubstantial claim
  • May combine two claims.

    Interpretationreasoning
    § 2.3. Inference · p. 4

    YOLO is extremely fast at test time since it only requires a single network evaluation, unlike classifier-based methods.

    “YOLO is extremely fast at test time since it only requires a single network evaluation, unlike classifier-based methods.”
    Quote in the paperSubstantial claim
  • May combine two claims.

    Premisereasoning
    § 4.5. Generalizability: Person Detection in Artwork · p. 7

    YOLO models the size and shape of objects, as well as relationships between objects and where objects commonly appear.

    “Like DPM, YOLO models the size and shape of objects, as well as relationships between objects and where objects commonly appear.”
    Quote in the paperSubstantial claim
  • May combine two claims.

    Premisecitation
    § 3. Comparison to Other Detection Systems · p. 5

    MultiBox cannot perform general object detection and is still just a piece in a larger detection pipeline.

    “However, MultiBox cannot perform general object detection and is still just a piece in a larger detection pipeline, requiring further image patch classification.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [8] [8] D. Erhan, C. Szegedy, A. Toshev, and D. Anguelov. Scalable object detection using deep neural networks. In Computer Vision and Pattern Recognition (CVPR), 2014 IEEE Conference on , pages 2155-2162. IEEE, 2014. 5, 6 - source
  • May combine two claims.

    Interpretationnone given
    § 3. Comparison to Other Detection Systems · p. 5

    YOLO is a general purpose detector that learns to detect a variety of objects simultaneously.

    “YOLO is a general purpose detector that learns to detect a variety of objects simultaneously.”
    Quote in the paperSubstantial claim
  • May combine two claims.

    Premisereasoning
    § 3. Comparison to Other Detection Systems · p. 5

    OverFeat cannot reason about global context and thus requires significant post-processing to produce coherent detections.

    “OverFeat cannot reason about global context and thus requires significant post-processing to produce coherent detections.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [32] [32] P. Sermanet, D. Eigen, X. Zhang, M. Mathieu, R. Fergus, and Y. LeCun. Overfeat: Integrated recognition, localization and detection using convolutional networks. CoRR , abs/1312.6229, 2013. 4, 5 -
  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given
    § 3. Comparison to Other Detection Systems · p. 5

    YOLO predicts both bounding boxes and class probabilities for multiple objects of multiple classes in an image.

    “YOLO predicts both bounding boxes and class probabilities for multiple objects of multiple classes in an image.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 2.1. Network Design · p. 2

    YOLO's network architecture is inspired by the GoogLeNet model for image classification.

    “Our network architecture is inspired by the GoogLeNet model for image classification [34].”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [34] [34] C. Szegedy, W. Liu, Y. Jia, P. Sermanet, S. Reed, D. Anguelov, D. Erhan, V. Vanhoucke, and A. Rabinovich. Going deeper with convolutions. CoRR , abs/1409.4842, 2014. 2 - source
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 3. Comparison to Other Detection Systems · p. 5

    MultiGrasp only needs to predict a single graspable region for an image containing one object.

    “MultiGrasp only needs to predict a single graspable region for an image containing one object.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [27] [27] J. Redmon and A. Angelova. Real-time grasp detection using convolutional neural networks. CoRR , abs/1412.3128, 2014. 5 - source
  • May combine two claims.

    Interpretationnone given
    § 2.4. Limitations of YOLO · p. 4

    YOLO imposes strong spatial constraints on bounding box predictions because each grid cell only predicts two boxes and can only have one class.

    “YOLO imposes strong spatial constraints on bounding box predictions since each grid cell only predicts two boxes and can only have one class.”
    Quote in the paperNot a substantial claim
  • May combine two claims.

    Interpretationreasoning
    § 2.2. Training · p. 3

    Each bounding box predictor gets better at predicting certain sizes, aspect ratios, or classes of object, improving overall recall.

    “Each predictor gets better at predicting certain sizes, aspect ratios, or classes of object, improving overall recall.”
    Quote in the paperSubstantial claim
  • May combine two claims.

    Interpretationreasoning
    § 2.2. Training · p. 3

    In every image many grid cells do not contain any object, pushing confidence scores towards zero and often overpowering the gradient from cells that do contain objects, which can lead to model instability and training divergence early on.

    “Also, in every image many grid cells do not contain any object. This pushes the 'confidence' scores of those cells towards zero, often overpowering the gradient from cells that do contain objects. This can lead to model instability, causing training to diverge early on.”
    Quote in the paperSubstantial claim
  • Classified as probably not a substantial claim.

    Premisecitation
    § 4.2. VOC 2007 Error Analysis · p. 6

    Fast R-CNN is one of the highest performing detectors on PASCAL.

    “We compare YOLO to Fast RCNN since Fast R-CNN is one of the highest performing detectors on PASCAL and it's detections are publicly available.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [14] [14] R. B. Girshick. Fast R-CNN. CoRR , abs/1504.08083, 2015. 2, 5, 6, 7 - source
  • Classified as probably not a substantial claim.

    Premisenone given
    § 4.2. VOC 2007 Error Analysis · p. 6

    Fast R-CNN's detections are publicly available.

    “We compare YOLO to Fast RCNN since Fast R-CNN is one of the highest performing detectors on PASCAL and it's detections are publicly available.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [14] [14] R. B. Girshick. Fast R-CNN. CoRR , abs/1504.08083, 2015. 2, 5, 6, 7 - source
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductcitation
    § 4.2. VOC 2007 Error Analysis · p. 6

    The error analysis uses the methodology and tools of Hoiem et al. [19].

    “We use the methodology and tools of Hoiem et al. [19]”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [19] [19] D. Hoiem, Y. Chodpathumwan, and Q. Dai. Diagnosing error in object detectors. In Computer Vision-ECCV 2012 , pages 340-353. Springer, 2012. 6 - source
  • Classified as probably not a substantial claim.

    Premisecitation
    § 4.5. Generalizability: Person Detection in Artwork · p. 7

    The Picasso Dataset and People-Art Dataset are for testing person detection on artwork.

    “We compare YOLO to other detection systems on the Picasso Dataset [12] and the People-Art Dataset [3], two datasets for testing person detection on artwork.”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [12] [12] S. Ginosar, D. Haas, T. Brown, and J. Malik. Detecting people in cubist art. In Computer Vision-ECCV 2014 Workshops , pages 101-116. Springer, 2014. 7 - source
    • cites [3] [3] H. Cai, Q. Wu, T. Corradi, and P. Hall. The crossdepiction problem: Computer vision algorithms for recognising objects in artwork and in photographs. arXiv preprint arXiv:1505.00110 , 2015. 7 - source
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation
    § 4.1. Comparison to Other Real-Time Systems · p. 6

    Faster R-CNN replaces selective search with a neural network to propose bounding boxes.

    “The recent Faster R-CNN replaces selective search with a neural network to propose bounding boxes, similar to Szegedy et al. [8]”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [8] [8] D. Erhan, C. Szegedy, A. Toshev, and D. Anguelov. Scalable object detection using deep neural networks. In Computer Vision and Pattern Recognition (CVPR), 2014 IEEE Conference on , pages 2155-2162. IEEE, 2014. 5, 6 - source
32 of 32
  • The paper relies on this without support.

    Academic datasets for object detection draw the training and testing data from the same distribution.

    “Academic datasets for object detection draw the training and testing data from the same distribution.”
  • The paper relies on this without support.

    Artwork and natural images are similar in terms of the size and shape of objects.

    “Artwork and natural images are very different on a pixel level but they are similar in terms of the size and shape of objects”
  • The paper relies on this without support.

    Selective Search is tuned for natural images.

    “R-CNN uses Selective Search for bounding box proposals which is tuned for natural images.”
  • The paper relies on this without support.

    The classifier step in R-CNN only sees small regions and needs good proposals.

    “The classifier step in R-CNN only sees small regions and needs good proposals.”
4 of 4

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?

  • YOLOv1 and YOLOv2 employed MSE/L2 loss for classification.

    “which was not thoroughly investigated in prevIoUs work such as YOLOv1 [20] and YOLOv2 [51] employing MSE/L2 loss (Mean Squared Error).”

    From Object Detection in 20 Years: A Survey (2023) · cites this paper as [20]

    Section 2.2 (Training, p. 3) says YOLO optimizes sum-squared error and that its loss penalizes classification error when an object is present. This supports the claim for YOLOv1, but this paper does not establish what loss YOLOv2 used. (LLM, conf=0.86)

    Experimental — accuracy not yet measured
  • YOLO series models v2-v5 are some of the most advanced object detection methods based on DNN, which have good performance in terms of speed and precision.

    “the YOLO series models v2-v5 [144-147] are some of the most advanced object detection methods based on DNN, which have good performance in terms of speed and precision.”

    From A survey of model compression strategies for object detection (2023) · cites this paper as [144]

    The Abstract (p. 1) and §4.1 (p. 5) describe the original YOLO as a neural-network detector with real-time speed and strong accuracy relative to other real-time detectors. The paper does not evaluate YOLO versions v2–v5, so it cannot substantiate the claim about those specific models. It also notes in the Introduction (p. 2) that its YOLO model lagged behind state-of-the-art detectors in accuracy. (LLM, conf=0.88)

    Experimental — accuracy not yet measured
  • Various deep learning-based intelligent services and applications have changed many aspects of people's lives due to DL advantages in computer vision and natural language processing.

    “As a typical and more widely used new form of applications [9], various deep learning-based intelligent services and applications have changed many aspects of people's lives due to the great advantages of Deep Learning (DL) in the fields of Computer Vision (CV) and Natural Language Processing (NLP) [10].”

    From Convergence of Edge Computing and Deep Learning: A Comprehensive Survey (2020) · cites this paper as [9]

    The Introduction (p. 1) describes potential applications of fast object detection in driving, assistive devices, and robotics, supporting the computer-vision portion of the claim. It does not address natural language processing or establish that deep-learning applications have already changed many aspects of people's lives. (LLM, conf=0.86)

    Experimental — accuracy not yet measured
  • YOLO's CNN predicts bounding boxes and class probabilities for all objects in an image.

    “A convolutional neural network predicts the bounding boxes as well as class probabilities for all the objects depicted in an image.”
  • YOLO reframes object detection as a regression problem rather than classification.

    “Authors of YOLO [56] have reframed the problem of object detection as a regression problem instead of classification problem.”

Cites

Flagged first, then by citations

Research works printed but not linked: 1

  1. mishkin, 0, Models accuracy on imagenet 2012 val

Cited by

Sorted by citations

Authors
Redmon, J., Divvala, S., Girshick, R., & Farhadi, A.
Affiliations
  • University of Washington (US)
  • Meta (United States) (US)
Venue
2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
Publisher
IEEE
Published
2016-06
Pages
779-788
ISSN
1063-6919 (ISSN-L)
OpenAlex
W1483870316
Cited by
53,803
References
42
Citations by year
2026: 5347 · 2025: 8912 · 2024: 8953 · 2023: 8264 · 2022: 6785 · 2021: 5741 · 2020: 4223 · 2019: 3286 · 2018: 1560 · 2017: 635 · 2016: 90 · 2015: 4 · 2014: 1
FWCI
644.23
Type
conference paper
Language
en
Open access
Locations
  • IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • arXiv (Cornell University)
  • Zenodo (CERN European Organization for Nuclear Research) (cc-by)
Funders
  • National Science Foundation (1338054)
  • Office of Naval Research (N00014-13-1-0720, N00014)
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, YOLO, real-time detection, bounding box regression, end-to-end learning, false positive reduction, domain generalization

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.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-10.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-10.

Cited sources are checked by gpt-6-sol.

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