The DOI matches its Crossref record; 106 of the 111 references it cites were found in a registry or an index.
Multiscale Vision Transformers
Open accessAbstract
We present Multiscale Vision Transformers (MViT) for video and image recognition, by connecting the seminal idea of multiscale feature hierarchies with transformer models. Multiscale Transformers have several channel-resolution scale stages. Starting from the input resolution and a small channel dimension, the stages hierarchically expand the channel capacity while reducing the spatial resolution. This creates a multiscale pyramid of features with early layers operating at high spatial resolution to model simple low-level visual information, and deeper layers at spatially coarse, but complex, high-dimensional features. We evaluate this fundamental architectural prior for modeling the dense nature of visual signals for a variety of video recognition tasks where it outperforms concurrent vision transformers that rely on large scale external pre-training and are 5-10× more costly in computation and parameters. We further remove the temporal dimension and apply our model for image classification where it outperforms prior work on vision transformers. Code is available at: https://github.com/facebookresearch/SlowFast.
Overview
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
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
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 2×
“As in original ViT [25], we use residual connections [45] and Layer Normalization (LN) [2] in the pre-normalization configuration that applies LN at the beginning of the residual function, and our MLPs consist of two linear layers with GELU activation [48], where the first layer expands the dimension from D to 4D, and…”
MethodResultVaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404 · cited 6×
“In contrast to original Multi Head Attention (MHA) operators [98], where the channel dimension and the spatiotemporal resolution remains fixed, MHPA pools the sequence of latent tensors to reduce the sequence length (resolution) of the attended input.”
BackgroundMethodInfluentialVery Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148 · cited 1×
“Incorporating downsampling, shift invariance, and shared weights, ConvNets are de-facto standard backbones for computer vision tasks for image [65, 62, 88, 90, 46, 12, 15, 34, 93, 81, 41] and video [87, 31, 11, 79, 69, 106, 96, 30, 105, 35, 29, 117, 57].”
BackgroundImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155 · cited 1×
“Incorporating downsampling, shift invariance, and shared weights, ConvNets are de-facto standard backbones for computer vision tasks for image [70, 67, 94, 96, 51, 15, 18, 39, 99, 87, 46] and video [93, 36, 14, 85, 74, 112, 102, 34, 111, 40, 33, 123, 62].”
BackgroundFaster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
Ren, S., He, K., Girshick, R., & Sun, J. (2017) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2016.2577031 · Cited by 73,472
Faster 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
ImageNet: 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
Going deeper with convolutions
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298594 · Cited by 48,562
Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., Lin, S., & Guo, B. (2021) · 2021 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.00986 · Cited by 35,534 · cited 1×
“Finally, compared to the best model of concurrent Swin [79] (which was designed for image recognition), MViT has +0.”
ResultRethinking the Inception Architecture for Computer Vision
Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., & Wojna, Z. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.308 · Cited by 32,138
Research works printed but not linked: 11
- gao, 2019, Res2net: A new multi-scale backbone architecture, IEEE PAMI
- fan, 2020, Pyslowfast
- feichtenhofer, 2019, SlowFast networks for video recognition in ActivityNet challenge 2019
- 10.1109/CVPR.2009.5206848 · DOI · Paper page
- 10.1109/CVPR.2019.00033 · DOI
- hang, 2020, ResNeSt Split-Attention Networks
- 10.1109/ICCV.2019.00718 · DOI
- 10.1109/CVPR.2015.7298594 · DOI · Paper page
- carion, 2020, Endto-end object detection with transformers, Proc ECCV, 213
- loshchilov, 2018, Fixing weight decay regularization in adam
- qiu, 2017, Learning spatiotemporal representation with pseudo-3d residual networks, Proc ICCV
Cited by
Sorted by citations
Items 201–300 of 1,003
Look Outside the Room: Synthesizing A Consistent Long-Term 3D Scene Video from A Single Image
Ren, X., & Wang, X. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.00355 · Cited by 48
Multi-scale hybrid vision transformer and Sinkhorn tokenizer for sewer defect classification
Haurum, J. B., Madadi, M., Escalera, S., & Moeslund, T. B. (2022) · Automation in Construction · doi:10.1016/j.autcon.2022.104614 · Cited by 46
“This problem has been approached in two ways, by introducing either hierarchical representations inspired by classical CNN architecture design [53 – 56] or multi-scale representations by applying different ViTs sequentially [59] or working on variations of the input in parallel [60,58].”
MethodSkeleTR: Towards Skeleton-based Action Recognition in the Wild
Duan, H., Xu, M., Shuai, B., Modolo, D., Tu, Z., Tighe, J., & Bergamo, A. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.01254 · Cited by 48
“With the great success in NLP [54, 13], Transformers have been increasingly applied to di-versified vision tasks [14, 34, 5, 74] as well as the RGB-based action understanding [35, 18, 1, 2, 63, 71], and have achieved promising results.”
BackgroundMGMAE: Motion Guided Masking for Video Masked Autoencoding
Huang, B., Zhao, Z., Zhang, G., Qiao, Y., & Wang, L. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.01241 · Cited by 52
MHST: Multiscale Head Selection Transformer for Hyperspectral and LiDAR Classification
Ni, K., Wang, D., Zheng, Z., & Wang, P. (2024) · IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing · doi:10.1109/jstars.2024.3366614 · Cited by 43
“Multiple layers of conventional global transformers [57] are utilized to integrate multimodal features from local spectral space f cnn and are performed initial global spectral feature extraction of AGFE.”
MethodWang, H., Liu, F., Jiao, L., Wang, J., Hao, Z., Li, S., Li, L., Chen, P., & Liu, X. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v38i6.28347 · Cited by 53
Multiscaled Multi-Head Attention-Based Video Transformer Network for Hand Gesture Recognition
Garg, M., Ghosh, D., & Pradhan, P. M. (2023) · IEEE Signal Processing Letters · doi:10.1109/lsp.2023.3241857 · Cited by 47
“Further, a transformer model that extracts multiscaled features using pooling in Multiscale Vision Transformers (MViT) was proposed in [10].”
MethodConv-trans dual network for landslide detection of multi-channel optical remote sensing images
Chen, X., Liu, M.-Z., Li, D.-F., Jia, J., Yang, A., Zheng, W.-F., & Yin, L. (2023) · Frontiers in Earth Science · doi:10.3389/feart.2023.1182145 · Cited by 41
Masked Motion Encoding for Self-Supervised Video Representation Learning
Sun, X., Chen, P., Chen, L., Li, C., Li, T. H., Tan, M., & Gan, C. (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/cvpr52729.2023.00222 · Cited by 48
“1 MViT [26] MViT-B K400 ! 100 64 455× 1× 3 37M 67.”
MethodMahmood, T., Wahid, A., Hong, J. S., Kim, S. G., & Park, K. R. (2024) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2024.108824 · Cited by 43
Stargazer: A Transformer-based Driver Action Detection System for Intelligent Transportation
Liang, J., Zhu, H., Zhang, E., & Zhang, J. (2022) · 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw56347.2022.00356 · Cited by 37
“Our model is based on the improved multi-scale vision transformers (MViT v2) [11, 19], which learn a hierarchy from dense (in”
BackgroundMethodInfluentialImplicit Temporal Modeling with Learnable Alignment for Video Recognition
Tu, S., Dai, Q., Wu, Z., Cheng, Z.-Q., Hu, H., & Jiang, Y. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.01825 · Cited by 52
“Various approaches have been proposed, including convolution-based * methods [49, 55, 53, 10, 60, 18, 17, 76] and transformer-based methods [3, 5, 20, 15, 31, 36, 50, 61].”
BackgroundMethodHyperspectral and LiDAR Classification via Frequency Domain-Based Network
Ni, K., Wang, D., Zhao, G., Zheng, Z., & Wang, P. (2024) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2024.3435793 · Cited by 40
“To reduce the computational cost of the transformer, Fan et al. [49] inserted a multihead pooling attention mechanism into the transformer to reduce the spatial dimension and facilitate the efficient representation of complex visual data.”
BackgroundLearning representational invariances for data-efficient action recognition
Zou, Y., Choi, J., Wang, Q., & Huang, J. (2022) · Computer Vision and Image Understanding · doi:10.1016/j.cviu.2022.103597 · Cited by 43
BoxeR: Box-Attention for 2D and 3D Transformers
Nguyen, D.-K., Ju, J., Booij, O., Oswald, M. R., & Snoek, C. G. M. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.00473 · Cited by 41
“As the ViT becomes computationally more expensive with high resolution images, while only producing a single-scale feature map, several works [8,24] have focused on speeding up the self-attention computation and generating multi-scale feature maps for object detection and segmentation.”
MethodDynamic Temporal Filtering in Video Models
Long, F.-C., Qiu, Z.-F., Pan, Y.-W., Yao, T., Ngo, C., & Mei, T. (2022) · European Conference on Computer Vision · doi:10.1007/978-3-031-19833-5_28 · Cited by 29
“Similarly, the popularity of image Transformer leads to the investigation of the video Transformer architectures [1,2,8,29,31].”
BackgroundSTMixer: A One-Stage Sparse Action Detector
Tao, W., Cao, M., Gao, Z., Wu, G., & Wang, L. (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/cvpr52729.2023.01414 · Cited by 38
“SlowFast [11] and MViT [9] directly use the RoI features for action classification.”
BackgroundMethodCATNet: Convolutional attention and transformer for monocular depth estimation
Tang, S., Lu, T., Liu, X., Zhou, H., & Zhang, Y. (2023) · Pattern Recognition · doi:10.1016/j.patcog.2023.109982 · Cited by 38
Disentangling Spatial and Temporal Learning for Efficient Image-to-Video Transfer Learning
Qing, Z., Zhang, S., Huang, Z., Zhang, Y., Gao, C., Zhao, D., & Sang, N. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.01281 · Cited by 41
“Due to the limited receptive field of convolutional networks, Transformer-based approaches [1, 3, 13, 36, 53, 32, 40, 47, 43, 4] with global attention have achieved promising performance.”
BackgroundIbh, M., Graßhof, S., Witzner, D., & Madeleine, P. (2023) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw59228.2023.00548 · Cited by 40
“Including vision transformers [1, 12, 18, 22] used for HAR.”
An Efficient Spatio-Temporal Pyramid Transformer for Action Detection
Weng, Y., Pan, Z., Han, M., Chang, X., & Zhuang, B. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-19830-4_21 · Cited by 41
GestFormer: Multiscale Wavelet Pooling Transformer Network for Dynamic Hand Gesture Recognition
Garg, M., Ghosh, D., & Pradhan, P. M. (2024) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw63382.2024.00254 · Cited by 42
“Pyramid Vision Transformer (PvT) [54], Pyramid Pooling Transformer (P2T) [57], MsMHA-VTN [24], MViT [22], Improved MViT [36], PSViT [9], POSTER [67] are some methods which use pooling to reduce the sequence length and reduce the computation cost of the entire system.”
MethodIntegrally Pre-Trained Transformer Pyramid Networks
Tian, Y., Xie, L., Wang, Z., Wei, L., Zhang, X., Jiao, J., Wang, Y., Tian, Q., & Ye, Q. (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/cvpr52729.2023.01785 · Cited by 39
“…such as introducing convolution into the transformer architecture so that the relationship between neighborhood tokens is better formulated [16,24,40,47,59,60], interacting between hybrid information [51], using window [20, 45, 74] or local [71] self-attentions to replace global…”
BackgroundHybrid dilated multilayer faster RCNN for object detection
Xin, F., Zhang, H., & Pan, H. (2023) · The Visual Computer · doi:10.1007/s00371-023-02789-y · Cited by 39
TranPhys: Spatiotemporal Masked Transformer Steered Remote Photoplethysmography Estimation
Shao, H., Luo, L., Qian, J., Chen, S., Hu, C., & Yang, J. J. (2023) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2023.3307700 · Cited by 36
“Inspired by the low-to-high-level visual deep modeling [76], we improve the traditional ViT multi-head self-attention into a form of targeted temporal pooling mechanism.”
MethodInterFormer Real-time Interactive Image Segmentation
You, H., Yang, H., Sun, K., Zhang, S., Cao, L., Jiang, G., & Ji, R. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.02038 · Cited by 36
“This inspired a series of ViT-based works [47, 56, 49, 14, 39, 51, 24, 23] that have successfully tackled various computer vision tasks.”
BackgroundScalable Video Object Segmentation With Identification Mechanism
Yang, Z., Miao, J., Wei, Y., Wang, W., Wang, X., & Yang, Y. (2024) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2024.3383592 · Cited by 49
“Other methods [77], [78] have sought to minimize the redundancy of spatial information by constructing multi-scale vision transformers for video processing.”
MethodPSLT: A Light-Weight Vision Transformer With Ladder Self-Attention and Progressive Shift
Wu, G., Zheng, W., Lu, Y., & Tian, Q. C. (2023) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2023.3265499 · Cited by 36
“Recent transformer architectures [13], [14], [15], [37], [41] have adopted hierarchical structures like popular CNNs to yield multi-scale features through patch merging to progressively reduce the”
BackgroundResultRe2TAL: Rewiring Pretrained Video Backbones for Reversible Temporal Action Localization
Zhao, C., Liu, S., Mangalam, K., & Ghanem, B. (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/cvpr52729.2023.01025 · Cited by 40
“Leveraging long-term temporal connection via the attention mechanism, transformers have quickly gain popularity as the backbone of choice for several video recognition workloads [3, 5, 15, 32, 45, 70].”
BackgroundSelf-citationLatency Matters: Real-Time Action Forecasting Transformer
Girase, H., Agarwal, N., Choi, C., & Mangalam, K. (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/cvpr52729.2023.01799 · Cited by 36
“For feature extraction, we fix the pretrained 16x4 (K400+IN1K) and 32x3 (K700) MViT-B backbones [12, 49] trained for recognition on the EK100 Model Setting Top-5 Recall AT FL SCM 2×AT”
BackgroundMethodInfluentialSelf-citationSwinWave-SR: Multi-scale lightweight underwater image super-resolution
Dharejo, F. A., Ganapathi, I. I., Zawish, M., Alawode, B., Alathbah, M. A., Werghi, N., & Javed, S. (2023) · Information Fusion · doi:10.1016/j.inffus.2023.102127 · Cited by 37
CTNet: convolutional transformer network for diabetic retinopathy classification
Bala, R., Sharma, A., & Goel, N. (2023) · Neural Computing and Applications · doi:10.1007/s00521-023-09304-3 · Cited by 35
“Some variants of ViT use different patch sizes to bring multi-scale features [63, 64].”
A review of surrogate-assisted design optimization for improving urban wind environment
Wu, Y., & Quan, S. J. (2024) · Building and Environment · doi:10.1016/j.buildenv.2023.111157 · Cited by 37
Change3D: Revisiting Change Detection and Captioning from A Video Modeling Perspective
Zhu, D., Huang, X., Huang, H., Zhou, H., & Shao, Z. (2025) · 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/cvpr52734.2025.02236 · Cited by 37
“Recently, several works [2, 18, 32, 40, 43, 54] have explored the potential of Vision Transformer [17] for spatiotemporal learning, verifying the outstanding ability to capture long-term dependencies.”
Hutchinson, M., & Gadepally, V. N. (2021) · IEEE Access · doi:10.1109/access.2021.3115476 · Cited by 46
“information as a sequence of tokens for the transformer, VTN [217] which is based off a transformer model that processes long sequences of tokens, ViViT [218] which uses another pure-transformer architecture, and MViT [219] which introduce resolution and channel scaling in combination with the vision transformer.”
MethodPoint2Vec for Self-supervised Representation Learning on Point Clouds
Zeid, K. A., Schult, J., Hermans, A., & Leibe, B. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-54605-1_9 · Cited by 42
Multiscale Audio Spectrogram Transformer for Efficient Audio Classification
Zhu, W., & Omar, M. (2023) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp49357.2023.10096513 · Cited by 34
“Multiscale Transformers [35, 36] adopts several channel-resolution scale stages and hierarchically expands the channel capacity while reducing the spatial resolution.”
ARMBench: An Object-centric Benchmark Dataset for Robotic Manipulation
Mitash, C., Wang, F., Lu, S., Terhuja, V., Garaas, T. W., Polido, F., & Nambi, M. (2023) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra48891.2023.10160846 · Cited by 34
“MViT-B [26] count 2,951 2,165 4,070 recall 0.”
ResFormer: Scaling ViTs with Multi-Resolution Training
Tian, R., Wu, Z., Dai, Q., Hu, H., Qiao, Y., & Jiang, Y. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.02176 · Cited by 42
“In contrast, vanilla ViTs are equipped with tokens of fixed-dimension, other related attempts lay emphasis on multi-scale spatial dimension of features instead of input [18, 20] or perform in an unsupervised way [42], yet limited effort has been made to explore multi-resolution supervised training for ViTs.”
BackgroundMATIS: Masked-Attention Transformers for Surgical Instrument Segmentation
Ayobi, N., Pérez-Rondón, A., Rodríguez, S., & Arbeláez, P. (2023) · IEEE International Symposium on Biomedical Imaging (ISBI) · doi:10.1109/isbi53787.2023.10230819 · Cited by 34
“Moreover, video transformers have also achieved state-of-the-art results, again outperforming CNN-based methods [31, 32].”
MethodU-Flow: A U-Shaped Normalizing Flow for Anomaly Detection with Unsupervised Threshold
Tailanián, M., Pardo, Á., & Musé, P. (2024) · Journal of Mathematical Imaging and Vision · doi:10.1007/s10851-024-01193-y · Cited by 40
“Despite its simplicity, an ablation study presented in Section 5.2 shows that this combination strategy leads to better results than using each one of the Transformers alone, far better than other ResNet variants, and even better than MViT2.”
MethodInfluentialFT-TDR: Frequency-Guided Transformer and Top-Down Refinement Network for Blind Face Inpainting
Wang, J., Chen, S., Wu, Z., & Jiang, Y. (2022) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2022.3146774 · Cited by 34
“It is widely recognized that transformer-based architecture has recently demonstrated superior performance on a broad range of vision and language tasks [9], [10], [11], [12], mostly because of its strong capabilities in modeling long-range relation.”
BackgroundIt's Time for Artistic Correspondence in Music and Video
Surís, D., Vondrick, C., Russell, B. C., & Salamon, J. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.01031 · Cited by 35
“We append a [CLS] token to the input, to represent the full video.”
MethodSPARTAN: Self-supervised Spatiotemporal Transformers Approach to Group Activity Recognition
Chappa, N. V. S. R., Nguyen, P., Nelson, A., Seo, H., Li, X., Dobbs, P. D., & Luu, K. (2023) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw59228.2023.00544 · Cited by 37
“In the video domain, many works [5, 9, 17, 22, 34, 35, 44, 48, 50] exploited spatial and temporal self-attention to learn video representation efficiently.”
BackgroundMEFood: A Large-Scale Representative Benchmark of Quotidian Foods for the Middle East
Ansari, M. Y., & Qaraqe, M. (2023) · IEEE Access · doi:10.1109/access.2023.3234519 · Cited by 33
“that is popularly used in vision transformers [11], [19] and modern convolutional networks can be tuned to identify fine-grain ingredients.”
BackgroundSlovo: Russian Sign Language Dataset
Kapitanov, A., Kvanchiani, K., Nagaev, A., & Elizaveta, P. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-44137-0_6 · Cited by 33
“Multiscale Vision Transformer (MViT) [6] model was specifically designed for video recognition tasks and provides a significant performance”
MethodDynamic attention augmented graph network for video accident anticipation
Song, W.-F., Li, S., Chang, T., Xie, K.-P., Hao, A.-M., & Qin, H. (2023) · Pattern Recognition · doi:10.1016/j.patcog.2023.110071 · Cited by 38
Hierarchical compositional representations for few-shot action recognition
Li, C., Zhang, J., Wu, S., Jin, X., & Shan, S. (2024) · Computer Vision and Image Understanding · doi:10.1016/j.cviu.2023.103911 · Cited by 33
Learning Correlation Structures for Vision Transformers
Kim, M., Seo, P. H., Schmid, C., & Cho, M. (2024) · 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/cvpr52733.2024.01792 · Cited by 36
“For the sake of simplicity in derivation, ConvSA (Eqs.8-10) in Sec.3.3 is described as sharing channel-wise convolution weights across channels for key and value projection, which is not exactly the same as those used in previous ConvSA-based methods [2, 9, 15].”
InfluentialX-VARS: Introducing Explainability in Football Refereeing with Multi-Modal Large Language Models
Held, J., Itani, H., Cioppa, A., Giancola, S., Ghanem, B., & Droogenbroeck, M. V. (2024) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw63382.2024.00332 · Cited by 46
“Held et al . used the MViT [17, 36] video encoder to extract spatio-temporal features.”
MethodMulti-Scale CNN-Transformer Hybrid Network for Rail Fastener Defect Detection
He, J., Wang, W., Lv, F., Luo, H., Zhang, G., & Chen, Z. (2025) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2025.3540846 · Cited by 36
“MViT [22] incorporated multi-scale feature pyramids into the Transformer architecture, resulting in the multi-scale vision transformers model, which offers a more efficient solution for vision transformers.”
MorphMLP: An Efficient MLP-Like Backbone for Spatial-Temporal Representation Learning
Zhang, D. J., Li, K., Wang, Y., Chen, Y., Chandra, S., Qiao, Y., Liu, L., & Shou, M. Z. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-19833-5_14 · Cited by 44
“Compared with the-state-of-art method MViT[16], which is pre-trained on large video dataset Kinetics-600, our MorphMLP only pre-trained on ImageNet-1K can obtain better performance (70.”
BackgroundMethodInfluentialVideo Test-Time Adaptation for Action Recognition
Lin, W., Mirza, M. J., Koziński, M., Possegger, H., Kuehne, H., & Bischof, H. (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/cvpr52729.2023.02198 · Cited by 41
“Transformer-based models have also been applied for video recognition [1, 3, 8, 28, 32, 33, 56].”
MethodConvolutional Embedding Makes Hierarchical Vision Transformer Stronger
Wang, C., Xu, H., Zhang, X., Wang, L., Zheng, Z., & Liu, H. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-20044-1_42 · Cited by 38
CRFormer: Cross-Resolution Transformer for segmentation of grape leaf diseases with context mining
Zhang, X., Cen, C., Li, F., Liu, M., & Mu, W. (2023) · Expert Systems with Applications · doi:10.1016/j.eswa.2023.120324 · Cited by 32
An Overview of Vision Transformers for Image Processing: A Survey
Kameswari, C. S., Kavitha, J., Reddy, T. S., Chinthaguntla, B., Jagatheesaperumal, S. K., Gaftandzhieva, S. N., & Doneva, R. Z. (2023) · International Journal of Advanced Computer Science and Applications · doi:10.14569/ijacsa.2023.0140830 · Cited by 38
“Fan et al. [116] introduce Multi-scale Vision Transformers (MViT) for video and image recognition by fusing the foundational concept of multi-scale feature hierarchies with transformer models.”
BackgroundTransferring Vision-Language Models for Visual Recognition: A Classifier Perspective
Wu, W., Sun, Z., Song, Y., Wang, J., & Ouyang, W. (2023) · International Journal of Computer Vision · doi:10.1007/s11263-023-01876-w · Cited by 36
“…al., 2017; Xie et al., 2018; Tran et al., 2018; Wu et al., 2021a,b). ii) The Vision Transformer (ViT) based architectures for image (Dosovitskiy et al., 2020; Han et al., 2021; Liu et al., 2021) or video recognition (Berta-sius et al., 2021; Arnab et al., 2021; Liu et al., 2022; Fan et al., 2021).”
Multiscale Global Prompt Transformer for EEG-Based Driver Fatigue Recognition
Zhao, P., Lian, C., Xu, B., & Zeng, Z. (2024) · IEEE Transactions on Automation Science and Engineering · doi:10.1109/tase.2024.3382892 · Cited by 32
End-to-End Spatio-Temporal Action Localisation with Video Transformers
Gritsenko, A. A., Xiong, X., Djolonga, J., Dehghani, M., Sun, C., Lučić, M., Schmid, C., & Arnab, A. (2024) · 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/cvpr52733.2024.01739 · Cited by 40
“Another architectural baseline that we can compare to is that of a two-stage Fast-RCNN model using external person detections from [61], as used by [3, 13, 15, 62].”
MethodLearned Queries for Efficient Local Attention
Arar, M., Shamir, A., & Bermano, A. H. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.01057 · Cited by 37
“These backbones follow a pyramid structure similar to convolutional neural networks (CNNs) [8, 15, 58, 70], and process high-resolution inputs by restricting the self-attention to smaller windows, preferably with some overlap [55] or other forms of intercommunication [8, 35, 66].”
BackgroundIntegrally Migrating Pre-trained Transformer Encoder-decoders for Visual Object Detection
Liu, F., Zhang, X., Peng, Z., Guo, Z., Wan, F., Ji, X., & Ye, Q. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00628 · Cited by 38
“Vision transformers including ViT [6], Swin [25], MViT [7], and PvT [36] became promising models for image recognition.”
BackgroundPolyViT: Co-training Vision Transformers on Images, Videos and AudioPreprint
Likhosherstov, V., Arnab, A., Choromański, K., Lučić, M., Tay, Y., Weller, A., & Dehghani, M. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2111.12993 · Cited by 66
VPTR: Efficient Transformers for Video Prediction
Ye, X., & Bilodeau, G.-A. (2022) · Proceedings - International Conference on Pattern Recognition/Proceedings/International Conference on Pattern Recognition · doi:10.1109/icpr56361.2022.9956707 · Cited by 32
“patches, either 2D or 3D, and then tokenize the local patch by concatenation or Pooling [30].”
MethodQiu, C.-J., Song, Y.-Q., Liu, Y., Zhu, Y., Han, K., Sheng, V. S., & Liu, Z. (2024) · Biomedical Signal Processing and Control · doi:10.1016/j.bspc.2023.105827 · Cited by 31
Zhou, S., & Zhou, H. (2024) · Remote Sensing · doi:10.3390/rs16132416 · Cited by 34
Real-time accident anticipation for autonomous driving through monocular depth-enhanced 3D modeling
Liao, H., Li, Y., Li, Z., Bian, Z., Lee, J., Cui, Z., Zhang, G., & Xu, C. (2024) · Accident Analysis & Prevention · doi:10.1016/j.aap.2024.107760 · Cited by 31
“For instance, UniFormerv2 [37], VideoSwin [38], CAVG [39], and MVITv2 [40] all employ the transformer-based model to delineate the relative positioning of traffic agents such as vehicles and pedestrians, underscoring the potential of Transformers in enhancing model understanding of dynamic traffic…”
BackgroundSelf-citationDrone-HAT: Hybrid Attention Transformer for Complex Action Recognition in Drone Surveillance Videos
Khan, M., Ahmad, J., Saddik, A. E., Gueaieb, W., Masi, G., & Karray, F. (2024) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw63382.2024.00474 · Cited by 32
“For example, MViT combines multi-scale hierarchies with Transformers using lower input/channel dimensions, progressively increasing capacity at lower resolutions [11,20].”
MethodVideoMamba: Spatio-Temporal Selective State Space Model
Park, J., Kim, H.-S., Ko, K., Kim, M., & Kim, C. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-72698-9_1 · Cited by 45
Panoramic Vision Transformer for Saliency Detection in 360$$^\circ $$ Videos
Yun, H., Lee, S., & Kim, G. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-19833-5_25 · Cited by 36
A Large-Scale Robustness Analysis of Video Action Recognition Models
Schiappa, M. C., Biyani, N., Kamtam, P., Vyas, S., Palangi, H., Vineet, V., & Rawat, Y. S. (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/cvpr52729.2023.01412 · Cited by 37
TempEE : Temporal–Spatial Parallel Transformer for Radar Echo Extrapolation Beyond Autoregression
Chen, S., Shu, T., Zhao, H. Q., Zhong, G., & Chen, X. (2023) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2023.3311510 · Cited by 31
“[35] devised a multiscale ViT structure, and Zhou et al.”
MethodOng, K. L., Lee, C. P., Lim, H. S., Lim, K. M., & Alqahtani, A. (2023) · IEEE Access · doi:10.1109/access.2023.3321122 · Cited by 31
In the Eye of Transformer: Global–Local Correlation for Egocentric Gaze Estimation and Beyond
Lai, B., Liu, M., Ryan, F., & Rehg, J. M. (2023) · International Journal of Computer Vision · doi:10.1007/s11263-023-01879-7 · Cited by 42
“Methods Cls Token Pooling Top1-Acc Top5-Acc Mean Cls Acc MViT [15] X 64.”
BackgroundMethodInfluentialEfficient Video Action Detection with Token Dropout and Context Refinement
Chen, L., Zhan, T., Song, Y., Wu, G., & Wang, L. (2023) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv51070.2023.00953 · Cited by 30
“Several newly transformer-based methods [10, 52, 45] apply a ViT variant backbone and achieve competitive results following the two-stage pipeline.”
MethodA Comprehensive Survey of Recent Transformers in Image, Video and Diffusion Models
Le, D. P. C., Wang, D., & Le, V.-T. (2024) · Computers, materials & continua/Computers, materials & continua (Print) · doi:10.32604/cmc.2024.050790 · Cited by 36
Dual Guidance Enabled Fuzzy Inference for Enhanced Fine-Grained Recognition
Chen, Q., He, F., Wang, G., Bai, X. P., Cheng, L., & Ning, X. (2024) · IEEE Transactions on Fuzzy Systems · doi:10.1109/tfuzz.2024.3427654 · Cited by 32
“To alleviate the single-scale limitations of traditional ViT-based methods, several model architectures have been proposed, such as CvT [11], MViT [12], MViTv2 [13], and Swin transformer [14].”
SVFAP: Self-Supervised Video Facial Affect Perceiver
Sun, L., Lian, Z., Wang, K., He, Y., Xu, M., Sun, H., Liu, B., & Tao, J. (2024) · IEEE Transactions on Affective Computing · doi:10.1109/taffc.2024.3436913 · Cited by 38
“After that, numerous ViT variants have emerged by incorporating more or less vision-friendly priors (e.g., local self-attention and hierarchical design) to reduce the quadratic scaling cost of vanilla ViT and improve model performance in vision tasks [56], [57], [58], [59], [60].”
Label Enhancement-Based Multiscale Transformer for Palm-Vein Recognition
Qin, H., Gong, C., Li, Y., Gao, X., & El‐Yacoubi, M. A. (2023) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2023.3261909 · Cited by 30
“These multiscale transformer-based approaches [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62], [63], [64] employed the self-attention mechanism to well model the global interaction among token embeddings, but they lacked a locality mechanism for local information extraction.”
MethodSelf-citation(2023) · doi:10.1109/cvpr52729.2023.00627 · Cited by 28
“We use MViT-B ( 16 × 4 ) [15] as the feature extractor for video modality, which is pre-trained on Kinetics-400 [37].”
MethodInfluentialSelf-citationKim, S., Park, H., & Park, S. (2024) · Biomedical Engineering Letters · doi:10.1007/s13534-024-00425-9 · Cited by 36
“Vision transformers have shown superior performances not only in many visual tasks [73–76], but also in medical imaging areas [77–80].”
BackgroundGroupMamba: Efficient Group-Based Visual State Space Model
Shaker, A., Wasim, S. T., Khan, S. A., Gall, J., & Khan, F. S. (2025) · 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/cvpr52734.2025.01389 · Cited by 45
“In recent years, the pioneering Vision Transformer (ViT) [10] has significantly impacted the computer vision field, including tasks such as image classification [12, 36, 38, 58], object detection [3, 44, 68, 71], and segmentation [5, 28, 51].”
Egocentric Early Action Prediction via Multimodal Transformer-Based Dual Action Prediction
Guan, W., Song, X., Wang, K., Wen, H., Ni, H., Wang, Y., & Chang, X. (2023) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2023.3248271 · Cited by 29
“The motivation for using Transformer lies in its powerful capability in the non-sequential relation modeling in many computer vision tasks [17], [18].”
BackgroundMethodZoom-VQA: Patches, Frames and Clips Integration for Video Quality Assessment
Zhao, K., Yuan, K., Sun, M. B., & Wen, X. (2023) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw59228.2023.00137 · Cited by 32
“, ViViT [1], MViT [9], Video Swin Transformer [28]).”
BackgroundFine-Grained Spatio-Temporal Parsing Network for Action Quality Assessment
Gedamu, K., Ji, Y., Yang, Y., Shao, J., & Shen, H. (2023) · IEEE Transactions on Image Processing · doi:10.1109/tip.2023.3331212 · Cited by 29
“1) Actor-Centric Multiscale Transformer: Inspired by the previous methods [58], the channel resolution of our spatiotemporal multiscale transformer varies at different stages of the network, which progressively increases channel dimensions.”
BackgroundMethodWang, C., Zhang, X., Yang, W., Wang, G., Li, X., Wang, J., & Lu, B. (2024) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2024.3385752 · Cited by 28
“Several approaches have been proposed to introduce multiscale information into vision transformer (VIT) by changing the patch size [29], [30], improving the multihead attention (MHA) [31], [32], or adding additional branches [33], [34].”
BackgroundHybrid ViT-CNN Network for Fine-Grained Image Classification
Shao, R., Bi, X., & Chen, Z. (2024) · IEEE Signal Processing Letters · doi:10.1109/lsp.2024.3386112 · Cited by 28
“This is denoted as: where x ∈ R The difference between the MIT module and the existing multi-scale modules [30], [31], [32], [33], [34], [35] is that SK units are used instead of the commonly-used feature pyramids to extract multi-scale information.”
MethodMatching Multi-Scale Feature Sets in Vision Transformer for Few-Shot Classification
Song, M., Yao, F., Zhong, G., Ji, Z., & Zhang, X. (2024) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2024.3435003 · Cited by 28
“MViT [22], also inspired by FPN [20], combines the concept of multi-scale with Transformer to capture more temporal information than ViT.”
DualFormer: Local-Global Stratified Transformer for Efficient Video Recognition
Liang, Y., Zhou, P., Zimmermann, R., & Yan, S. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-19830-4_33 · Cited by 31
Multi-Scale Efficient Graph-Transformer for Whole Slide Image Classification
Ding, S., Li, J., Wang, J., Ying, S., & Shi, J. (2023) · IEEE Journal of Biomedical and Health Informatics · doi:10.1109/jbhi.2023.3317067 · Cited by 32
“Inspired by the feature pyramid of images in CNNs, the MVTs have been designed to learn multi-scale visual representations from images [21], [22], [32], [33].”
BackgroundMethodSeeing in Flowing: Adapting CLIP for Action Recognition with Motion Prompts Learning
Wang, Q., Du, J., Yan, K., & Ding, S. (2023) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3581783.3612490 · Cited by 30
“During the past few years, the developments of deep neural networks [22, 23, 25, 30, 38, 39, 45] and current transformers [1, 2, 7, 12, 24, 26, 47] have achieved promising progress in action recognition.”
BackgroundZhao, Z., Chen, Z., Li, J., Xie, X., Chen, K., Wang, X., & Shi, G. (2023) · Neurocomputing · doi:10.1016/j.neucom.2023.126903 · Cited by 27
Liu, Y., Huang, Q., Li, R., Chen, X., Zhao, Z., Zhao, S., Zhu, Y., & Zhang, H. (2024) · IEEE Transactions on Vehicular Technology · doi:10.1109/tvt.2024.3390414 · Cited by 27
“In this article, motivated by the aforementioned discussions and the significant contributions of artificial intelligence studies [19], [20], [21], we propose an IoSI-based collaborative perception framework called Select2Col to more comprehensively take account of the spatial-temporal importance of…”
MethodPaulraj, S., & Vairavasundaram, S. (2025) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2024.109496 · Cited by 27
Pixel-wise recognition for holistic surgical scene understanding
Ayobi, N., Rodríguez, S., Pérez, A., Hernández, I., Aparicio, N., Dessevres, E., Peña, S., Santander, J., Caicedo, J. I., Fernández, N., et al. (2025) · Medical Image Analysis · doi:10.1016/j.media.2025.103726 · Cited by 29
Dynamic Spatio-Temporal Specialization Learning for Fine-Grained Action Recognition
Li, T., Foo, L. G., Ke, Q., Rahmani, H., Wang, A., Wang, J., & Liu, J. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-19772-7_23 · Cited by 27
Few-shot human–object interaction video recognition with transformers
Li, Q., Xie, X., Jin, Z., & Shi, G. (2023) · Neural Networks · doi:10.1016/j.neunet.2023.01.019 · Cited by 29
Overview of temporal action detection based on deep learning
Hu, K., Shen, C., Wang, T., Xu, K., Xia, Q., Xia, M., & Cai, C. (2024) · Artificial Intelligence Review · doi:10.1007/s10462-023-10650-w · Cited by 32
Learning Local and Global Temporal Contexts for Video Semantic Segmentation
Sun, G., Liu, Y., Ding, H., Wu, M., & Gool, L. V. (2024) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2024.3387326 · Cited by 28
Zheng, Z., Zhong, Y., Zhao, J. J., Ma, A., & Zhang, L. (2024) · ISPRS Journal of Photogrammetry and Remote Sensing · doi:10.1016/j.isprsjprs.2024.07.001 · Cited by 29
VicTR: Video-conditioned Text Representations for Activity Recognition
Kahatapitiya, K., Arnab, A., Nagran, A., & Ryoo, M. S. (2024) · 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/cvpr52733.2024.01755 · Cited by 26
“Whether it is efficient architecture variants ranging from CNNs [13, 31, 57] to Transformers [2, 5, 12], training schemes from fully-supervised [7, 14] to self-supervised [15, 47, 53] or data regimes from unimodal [64, 76] to multimodal [20, 39], the progress has been steady and exciting.”
Background
Metadata
- Authors
- Fan, H., Xiong, B., Mangalam, K., Li, Y., Yan, Z., Malik, J., & Feichtenhofer, C.
- Affiliations
- Meta (Israel) (IL)
- Meta (United States) (US)
- Venue
- 2021 IEEE/CVF International Conference on Computer Vision (ICCV)
- Publisher
- IEEE
- Published
- 2021-10
- Pages
- 6804-6815
- OpenAlex
- W3202698164
- Cited by
- 1,407
- References
- 111
- Citations by year
- 2026: 147 · 2025: 345 · 2024: 380 · 2023: 349 · 2022: 176 · 2021: 10
- FWCI
- 74.81
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- IEEE/CVF International Conference on Computer Vision (ICCV) (publishedVersion)
- arXiv (Cornell University)
- TIB Data Manager
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Human Pose and Action Recognition
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Keywords
- video recognition, image classification, feature pyramid, spatial resolution reduction, computational efficiency
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
