What is the best multi-stage architecture for object recognition?
Open accessThe DOI matches its Crossref record; 56 of the 57 references it cites were found in a registry or an index.
Abstract
In many recent object recognition systems, feature extraction stages are generally composed of a filter bank, a non-linear transformation, and some sort of feature pooling layer. Most systems use only one stage of feature extraction in which the filters are hard-wired, or two stages where the filters in one or both stages are learned in supervised or unsupervised mode. This paper addresses three questions: 1. How does the non-linearities that follow the filter banks influence the recognition accuracy? 2. does learning the filter banks in an unsupervised or supervised manner improve the performance over random filters or hardwired filters? 3. Is there any advantage to using an architecture with two stages of feature extraction, rather than one? We show that using non-linearities that include rectification and local contrast normalization is the single most important ingredient for good accuracy on object recognition benchmarks. We show that two stages of feature extraction yield better accuracy than one. Most surprisingly, we show that a two-stage system with random filters can yield almost 63% recognition rate on Caltech-101, provided that the proper non-linearities and pooling layers are used. Finally, we show that with supervised refinement, the system achieves state-of-the-art performance on NORB dataset (5.6%) and unsupervised pre-training followed by supervised refinement produces good accuracy on Caltech-101 (> 65%), and the lowest known error rate on the undistorted, unprocessed MNIST dataset (0.53%).
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Citations
Cites
Gradient-based learning applied to document recognition
Lecun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998) · Proceedings of the IEEE · doi:10.1109/5.726791 · Cited by 58,805 · cited 6×
“Results are presented on the well-known Caltech-101 dataset [7], on the NORB object dataset [15], and on the MNIST dataset of handwritten digits [14].”
BackgroundMethodInfluentialSelf-citationDistinctive Image Features from Scale-Invariant Keypoints
Lowe, D. G. (2004) · International Journal of Computer Vision · doi:10.1023/b:visi.0000029664.99615.94 · Cited by 55,589 · cited 10×
“SIFT uses a rectification followed by winner-take-all operation over orientation , which is an extreme form of normalization.”
MethodInfluentialHistograms of Oriented Gradients for Human Detection
Dalal, N., & Triggs, B. (2005) · 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) · doi:10.1109/cvpr.2005.177 · Cited by 33,584 · cited 3×
“Particular embodiments of the single-stage systems use SIFT features [19, 13], HoG [6], Geometric Blur [5], and models inspired by the architecture of the mammalian primary visual cortex [24], to mention a few.”
MethodReducing the dimensionality of data with neural networks
Hinton, G. E., & Salakhutdinov, R. R. (2006) · Science · doi:10.1126/science.1127647 · Cited by 20,928
Harrison's Principles of Internal Medicine.
(1988) · Annals of Internal Medicine · doi:10.7326/0003-4819-108-3-499_2 · Cited by 12,514
Beyond Bags of Features: Spatial Pyramid Matching for Recognizing Natural Scene Categories
Lazebnik, S., Schmid, C., & Ponce, J. (2006) · 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition - Volume 2 (CVPR'06) · doi:10.1109/cvpr.2006.68 · Cited by 7,916 · cited 12×
“The performance of our two-stage system is similar to that of the best single-stage systems based on SIFT and PMK-SVM, perhaps because the PM Kernel is conceptually similar to our feature extractions stage.”
BackgroundMethodInfluentialGreedy layer-wise training of deep networks
Bengio, Y., Lamblin, P., Popovici, D., & Larochelle, H. (2007) · Advances in Neural Information Processing Systems 19 · doi:10.7551/mitpress/7503.003.0024 · Cited by 4,841
Sparse coding with an overcomplete basis set: A strategy employed by V1?
Olshausen, B. A., & Field, D. J. (1997) · Vision Research · doi:10.1016/s0042-6989(97)00169-7 · Cited by 3,642
Linear spatial pyramid matching using sparse coding for image classification
Yang, J., Yu, K., Gong, Y., & Huang, T. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206757 · Cited by 2,828
Fei-Fei, L., Fergus, R., & Perona, P. (2005) · 2004 Conference on Computer Vision and Pattern Recognition Workshop · doi:10.1109/cvpr.2004.383 · Cited by 2,502
Cited by
Items 401–500 of 1,007
Traffic sign detection based on convolutional neural networks
Wu, Y., Liu, Y., Li, J., Liu, H., & Hu, X. (2013) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn.2013.6706811 · Cited by 107
“Recently, Convolutional Neural Network has been adopted in object recognition for its high accuracy [10] [11] [12] [13].”
MethodLearning Structure and Strength of CNN Filters for Small Sample Size Training
Keshari, R., Vatsa, M., Singh, R., & Noore, A. (2018) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr.2018.00974 · Cited by 115
Reparameterized full-waveform inversion using deep neural networks
He, Q., & Wang, Y. (2020) · Geophysics · doi:10.1190/geo2019-0382.1 · Cited by 125
Classification of Histology Sections via Multispectral Convolutional Sparse Coding
Zhou, Y., Chang, H., Barner, K. E., Spellman, P. T., & Parvin, B. (2014) · 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.2014.394 · Cited by 115
“1Limited by space, we refer readers to [13, 21] for detailed discussions on local contrast normalization.”
BackgroundMethodLe, Q. V. (2015) · Cited by 155
Qin, W., Wu, J., Han, F., Yuan, Y., Zhao, W., Ibragimov, B., Gu, J., & Xing, L. (2018) · Physics in Medicine & Biology · doi:10.1088/1361-6560/aabd19 · Cited by 106
“…of each convolution layer can be defined as: where f(·) is an element-wise non-linear transform, here we choose tangent functions tanh for each layer computation, because tanh can produce better a result than sigmoid that has gradient vanishing problem and computation burden (Jarrett et al., 2009).”
BackgroundLearning deep representations via extreme learning machines
Yu, W., Zhuang, F., He, Q., & Shi, Z. (2014) · Neurocomputing · doi:10.1016/j.neucom.2014.03.077 · Cited by 115
SAR Automatic Target Recognition Based on Euclidean Distance Restricted Autoencoder
Deng, S., Du, L., Li, C., Ding, J., & Liu, H. (2017) · IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing · doi:10.1109/jstars.2017.2670083 · Cited by 105
“Deep learning algorithms have achieved state-of-theart results in numerous problems [10] concerning large-scale data such as classification [11] and inference [12].”
BackgroundA Convolutional Subunit Model for Neuronal Responses in Macaque V1
Vintch, B., Movshon, J. A., & Simoncelli, E. P. (2015) · The Journal of Neuroscience · doi:10.1523/jneurosci.2815-13.2015 · Cited by 103
Zhang, M., Jiang, Y., Li, C., & Yang, F. (2020) · Biosystems Engineering · doi:10.1016/j.biosystemseng.2020.01.018 · Cited by 110
MicroNet: Improving Image Recognition with Extremely Low FLOPs
Li, Y., Chen, Y., Dai, X., Chen, D., Liu, M., Yuan, L., Liu, Z., Zhang, L., & Vasconcelos, N. M. (2021) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.00052 · Cited by 159
“Dynamic ReLU [5] adapts slopes and intercepts of two linear functions in ReLU [29, 17].”
BackgroundInfluentialWhat Do We Understand About Convolutional Networks?Preprint
Hadji, I., & Wildes, R. P. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1803.08834 · Cited by 165
“ut also increase their practical applicability. Moreover, current realizations of ConvNets require massive amounts of data for training [84,88,91] and design decisions made greatly impact performance [23,77]. Deeper theoretical understanding should lessen dependence on data-driven design. While empirical studies have…”
BackgroundMethodInfluentialBio-inspired computer vision: Towards a synergistic approach of artificial and biological vision
Medathati, N. V. K., Neumann, H., Masson, G. S., & Kornprobst, P. (2016) · Computer Vision and Image Understanding · doi:10.1016/j.cviu.2016.04.009 · Cited by 124
“…have also mentioned how these bio-inspired computational blocks (e.g., normalisation) can be re-used in a computer vision framework to improve image processing algorithms (e.g., statistical whitening and source separation ( Lyu and Simoncelli, 2009 ), pattern recognition ( Jarrett et al., 2009 )).”
BackgroundAn improved faster-RCNN model for handwritten character recognition
Albahli, S., Nawaz, M., Javed, A., & Irtaza, A. (2021) · Arabian Journal for Science and Engineering · doi:10.1007/s13369-021-05471-4 · Cited by 107
“Some works have reported an accuracy of 98% to 99% for hand-written digit recognition [29].”
BackgroundResultInfluentialDeep Learning of Representations: Looking Forward
Bengio, Y. (2013) · arXiv (Cornell University) · doi:10.48550/arxiv.1305.0445 · Cited by 99
“Before covering the unsupervised representation learning algorithms, we quickly review the basic principles behind supervised representation learning algorithms such as the good old multi-layer neural networks.”
BackgroundMethodClassification of Image using Convolutional Neural Network (CNN)
Hossain, M. A., & Sajib, M. S. A. (2019) · Global Journal of Computer Science and Technology · doi:10.34257/gjcstdvol19is2pg13 · Cited by 160
Deep Learning in Object Recognition, Detection, and Segmentation
Wang, X. (2016) · Foundations and Trends® in Signal Processing · doi:10.1561/2000000071 · Cited by 128
“[126] first learned features in each layer in a unsupervised way [14, 117, 79], and then directly applied a multi-scale convolutional neural network [74] to pedestrian detection.”
MethodSpecific Emitter Identification Using Convolutional Neural Network-Based IQ Imbalance Estimators
Wong, L. J., Headley, W. C., & Michaels, A. J. (2019) · IEEE Access · doi:10.1109/access.2019.2903444 · Cited by 99
Deep Learning with Hierarchical Convolutional Factor Analysis
Chen, B., Polatkan, G., Sapiro, G., Blei, D. M., Dunson, D. B., & Carin, L. (2013) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2013.19 · Cited by 107
“the authors of [1], [7], [31], [4], [32], considering Caltech 101”
BackgroundTowards Robust Human-Robot Collaborative Manufacturing: Multimodal Fusion
Liu, H., Fang, T., Zhou, T., & Wang, L. (2018) · IEEE Access · doi:10.1109/access.2018.2884793 · Cited by 116
“Max-pooling makes each feature map invariant to local translations in the input map, which also proved to be useful in CNN [57], [58].”
MethodDiagnosis of Plant Cold Damage Based on Hyperspectral Imaging and Convolutional Neural Network
Yang, W., Yang, C., Hao, Z., Xie, C., & Li, M. (2019) · IEEE Access · doi:10.1109/access.2019.2936892 · Cited by 97
Khan, H. U., Shah, P. M., Shah, M. A., Islam, S. U., & Rodrigues, J. J. P. C. (2020) · Computer Communications · doi:10.1016/j.comcom.2020.01.013 · Cited by 110
More trainable inception-ResNet for face recognition
Peng, S., Huang, H., Chen, W., Zhang, L., & Fang, W. (2020) · Neurocomputing · doi:10.1016/j.neucom.2020.05.022 · Cited by 108
Unsupervised Learning of Visual Structure using Predictive Generative Networks
Lotter, W. E., Kreiman, G., & Cox, D. (2015) · arXiv (Cornell University) · doi:10.48550/arxiv.1511.06380 · Cited by 79
“…to the random initial weights, from which the latent variables can already be decoded fairly well, which is expected given the empirical evidence and theoretical justifications for the success of random weights in neural networks (Jarrett et al., 2009; Pinto et al., 2009; Saxe et al., 2010).”
BackgroundFeature Learning With Matrix Factorization Applied to Acoustic Scene Classification
Bisot, V., Serizel, R., Essid, S., & Gaël Richard, A. (2017) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2017.2690570 · Cited by 95
“vations [67] are used in hidden layers and softmax activations are applied to the output layer.”
MethodConvolutional Neural Network Hyper-Parameters Optimization based on Genetic Algorithms
Loussaief, S., & Abdelkrim, A. (2018) · International Journal of Advanced Computer Science and Applications · doi:10.14569/ijacsa.2018.091031 · Cited by 110
“In the Enhanced E-CNN-MP proposed method, batch normalization layers are incorporated between convolutional layers and ReLU layers.”
BackgroundInfluential(2010) · Cited by 94
Cunefare, D., Fang, L., Cooper, R. F., Dubra, A., Carroll, J., & Farsiu, S. (2017) · Scientific Reports · doi:10.1038/s41598-017-07103-0 · Cited by 100
“This lowers the computational burden and improves robustness to small image distortions by increasing spatial invariance 57 .”
BackgroundMethodA multi-scene deep learning model for image aesthetic evaluation
Wang, N., Zhao, M., Wang, L., Huang, J., Cai, C., & Xu, X. (2016) · Signal Processing Image Communication · doi:10.1016/j.image.2016.05.009 · Cited by 100
“Deep learning is very successful in solving classical computer vision tasks [8-10,14-15] such as object recognition [16], image classification [17].”
BackgroundA Biologically Inspired Appearance Model for Robust Visual Tracking
Zhang, S., Lan, X., Yao, H., Zhou, H., Tao, D., & Li, X. (2016) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2016.2586194 · Cited by 94
“3) Supported by the biological studies [ 56], [57] and empirical evaluation [ 42], [43], non-linear operators such as whitening, rectification and normalization operato rs can also enhance the representation capability.”
BackgroundMethodA Cloud Detection Method for Landsat 8 Images Based on PCANet
Zi, Y., Xie, F., & Jiang, Z. (2018) · Remote Sensing · doi:10.3390/rs10060877 · Cited by 104
“PCANet is easy to be trained and adapted to different data and tasks, and can avoid the problem of parameter training in traditional CNN that consumes too much time and requires various ad hoc tricks [54,55].”
MethodRecognition of handwritten characters using local gradient feature descriptors
Surinta, O., Karaaba, M. F., Schomaker, L., & Wiering, M. (2015) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2015.07.017 · Cited by 94
“Furthermore, the convolutional neural network (CNN) (LeCun et al., 1998) is used as a feature extraction and classi fi cation technique, and the accuracy obtained is 99.47% (Jarrett et al., 2009).”
MethodInfluentialVehicle Color Recognition With Spatial Pyramid Deep Learning
Hu, C., Bai, X., Li, Q., Chen, P., Xue, G., & Mei, L. (2015) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2015.2430892 · Cited by 95
Improving patch-based scene text script identification with ensembles of conjoined networks
Gómez, L., Nicolaou, A., & Karatzas, D. (2017) · Pattern Recognition · doi:10.1016/j.patcog.2017.01.032 · Cited by 92
“The CVSI-2015 [38] dataset comprises pre-segmented words in ten scripts: English, Hindi, Bengali, Oriya, Gujrathi, Punjabi, Kannada, Tamil, Telegu, and Arabic.”
Robust Convolutional Neural Networks for Image Recognition
Hayder, M., Haider, A., & Naz, E. (2015) · International Journal of Advanced Computer Science and Applications · doi:10.14569/ijacsa.2015.061115 · Cited by 118
“Thus several proposed methods are proposed for variety of applications such as image recognition [13, 14, 15, 16, 8], object detection [17, 18, 19], scene labeling [20], segmentation [21, 22], and variety of other tasks [23, 24, 25].”
MethodLocally-connected and convolutional neural networks for small footprint speaker recognition
Chen, Y., Moreno, I. L., Sainath, T. N., Visontai, M., Álvarez, R., & Parada, C. (2015) · Interspeech (USB) · doi:10.21437/interspeech.2015-297 · Cited by 92
Modern approaches in deep learning for SAR ATR
Wilmanski, M., Kreucher, C., & Lauer, J. (2016) · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · doi:10.1117/12.2220290 · Cited by 87
Principles of Temporal Processing Across the Cortical Hierarchy
Himberger, K. D., Chien, H.-Y., & Honey, C. J. (2018) · Neuroscience · doi:10.1016/j.neuroscience.2018.04.030 · Cited by 88
Unsupervised learning of invariant representations
Anselmi, F., Leibo, J. Z., Rosasco, L., Mutch, J., Tacchetti, A., & Poggio, T. (2016) · Theoretical Computer Science · doi:10.1016/j.tcs.2015.06.048 · Cited by 88
“[12]).”
Comparison of Support Vector Machine and Softmax Classifiers in Computer Vision
Qi, X., Wang, T., & Liu, J. (2017) · International Conference on Mechanical, Control and Computer Engineering (ICMCCE) · doi:10.1109/icmcce.2017.49 · Cited by 90
Learning Discriminative and Shareable Features for Scene Classification
Zuo, Z. C., Wang, G., Shuai, B., Zhao, L., Yang, Q., & Jiang, X. (2014) · Lecture notes in computer science · doi:10.1007/978-3-319-10590-1_36 · Cited by 92
A Survey and Taxonomy of FPGA-based Deep Learning Accelerators
Blaiech, A. G., Ben Khalifa, K., Valderrama, C., Fernandes, M. A. C., & Bedoui, M. H. (2019) · Journal of Systems Architecture · doi:10.1016/j.sysarc.2019.01.007 · Cited by 88
NeuFlow: Dataflow vision processing system-on-a-chip
Pham, P.-H., Jelaca, D., Farabet, C., Martini, B., LeCun, Y., & Culurciello, E. (2012) · 2012 IEEE 55th International Midwest Symposium on Circuits and Systems (MWSCAS) · doi:10.1109/mwscas.2012.6292202 · Cited by 86
“These computations can only be done with a hierarchical (logarithmic) architecture with multiple stages of processing in series [2].”
BackgroundSelf-citationHybrid Deep Learning for Face Verification
Sun, Y. J., Wang, X., & Tang, X. (2015) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2015.2505293 · Cited by 97
“We evaluate three activation functions proposed in literature to replace the function f(·) in Equation 1, 2, and 4, that is, the ReLU max (0, ·) [32], the hyperbolic tangent tanh(·) [36], and the absolute value rectified hyperbolic tangent abs(tanh(·)) [29].”
MethodTensor-factorized neural networks
Chien, J., & Bao, Y.-T. (2017) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2017.2690379 · Cited by 104
“The convolution and max-pooling are performed in one stage or two stages [38] similar to one or two hidden layers in an NN and a TFNN.”
MethodTechnology Aware Training in Memristive Neuromorphic Systems for Nonideal Synaptic Crossbars
Chakraborty, I., Roy, D., & Roy, K. (2018) · IEEE Transactions on Emerging Topics in Computational Intelligence · doi:10.1109/tetci.2018.2829919 · Cited by 88
Gjoreski, M., Gams, M., Luštrek, M., Genc, P., Garbas, J.-U., & Hassan, T. (2020) · IEEE Access · doi:10.1109/access.2020.2986810 · Cited by 106
“…units in the previous layers; (ii) shared weights, namely, units in a layer are organized in groups and all units in the same group share the same set of weights [57], [58]; and (iii) spatial or temporal sampling, namely, if the input is shifted, the feature map output will also be shifted [55].”
BackgroundTowards cross-category knowledge propagation for learning visual concepts
Qi, G.-J., Aggarwal, C. C., Rui, Y., Tian, Q. C., Chang, S., & Huang, T. S. (2011) · 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.2011.5995312 · Cited by 88
Blind Deep S3D Image Quality Evaluation via Local to Global Feature Aggregation
Oh, H., Ahn, S., Kim, J., & Lee, S. (2017) · IEEE Transactions on Image Processing · doi:10.1109/tip.2017.2725584 · Cited by 87
Deep Learning for Microalgae Classification
Correa, I., Drews, P., Botelho, S., de Souza, M. S., & Tavano, V. M. (2017) · 2017 16th IEEE International Conference on Machine Learning and Applications (ICMLA) · doi:10.1109/icmla.2017.0-183 · Cited by 70
“Among deep learning models, CNNs has turned to be the standard for classification of visual inputs [20].”
MethodAn End-to-End Deep Learning Approach for Epileptic Seizure Prediction
Xu, Y.-K., Yang, J., Zhao, S.-Q., Wu, H. C. H., & Sawan, M. (2020) · International Conference on Artificial Intelligence Circuits and Systems · doi:10.1109/aicas48895.2020.9073988 · Cited by 69
“The proposed network adopts 1-dimensional (1D) kernel in the early-stage convolution and max-pooling layer to make use of the EEG signal redundancy in the time domain and preserve information in the spatial domain [25], [26].”
MethodZooplanktoNet: Deep convolutional network for zooplankton classification
Dai, J., Wang, R., Zheng, H., Ji, G., & Qiao, X. (2016) · OCEANS 2016 - Shanghai · doi:10.1109/oceansap.2016.7485680 · Cited by 89
A machine learning approach for efficient uncertainty quantification using multiscale methods
Chan, S., & Elsheikh, A. H. (2017) · Journal of Computational Physics · doi:10.1016/j.jcp.2017.10.034 · Cited by 83
“Some popular choices for σ i are the hyperbolic tangent, the sigmoid function, and more recently, the rectified linear unit (Jarrett et al., 2009).”
BackgroundVehicle Type Classification Using Unsupervised Convolutional Neural Network
Dong, Z., Pei, M., He, Y., Liu, T., Dong, Y., & Jia, Y. (2014) · 2014 22nd International Conference on Pattern Recognition · doi:10.1109/icpr.2014.39 · Cited by 82
“Several rectifying strategies including the absolute value and the positive part are tried in [24], and the results are similar.”
MethodResultScheduling the Operation of a Connected Vehicular Network Using Deep Reinforcement Learning
Atallah, R. F., Assi, C. M., & Khabbaz, M. (2018) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2018.2832219 · Cited by 89
“The first hidden layer of the neural network convolves G (Nmax × 4) filters and then applies a rectifier nonlinearity [41].”
MethodA robust modulation classification method using convolutional neural networks
Zhou, S., Yin, Z., Wu, Z., Chen, Y., Zhao, N., & Yang, Z. (2019) · EURASIP Journal on Advances in Signal Processing · doi:10.1186/s13634-019-0616-6 · Cited by 88
“Max pooling is used in this paper, which is an operation that reports the maximum output within a pooling window [26].”
MethodEfficient Optimization of Convolutional Neural Networks Using Particle Swarm Optimization
Yamasaki, T., Honma, T., & Aizawa, K. (2017) · IEEE International Conference on Multimedia Big Data · doi:10.1109/bigmm.2017.69 · Cited by 81
“Further, the recognition performance of CNNs has been reported to be strongly dependent on the network configuration [1].”
BackgroundNLProlog: Reasoning with Weak Unification for Question Answering in Natural Language
Weber, L., Minervini, P., Münchmeyer, J., Leser, U., & Rocktäschel, T. (2019) · Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/p19-1618 · Cited by 77
Khemchandani, Y., O’Hagan, S., Samanta, S., Swainston, N., Roberts, T. J., Bollegala, D., & Kell, D. B. (2020) · Journal of Cheminformatics · doi:10.1186/s13321-020-00454-3 · Cited by 80
Eom, Y. H., Chung, Y., Park, M., Hong, S. B., & Kim, M. S. (2021) · Energy · doi:10.1016/j.energy.2021.120542 · Cited by 81
Color image classification via quaternion principal component analysis network
Zeng, R., Wu, J., Shao, Z., Chen, Y., Chen, B., Senhadji, L., & Shu, H. (2016) · Neurocomputing · doi:10.1016/j.neucom.2016.08.006 · Cited by 89
“The key factor of the success of deep learning in image classification is the use of convolutional architecture [6-12].”
BackgroundRobust Visual Tracking via Basis Matching
Zhang, S., Lan, X., Qi, Y., & Yuen, P. C. (2016) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2016.2539860 · Cited by 81
“and empirical evaluation [38], [39], nonlinear operators such as whitening, rectification, and normalization operators can also enhance the representation ability.”
BackgroundMutual exclusivity loss for semi-supervised deep learning
Sajjadi, M., Javanmardi, M., & Tasdizen, T. (2016) · 2016 IEEE International Conference on Image Processing (ICIP) · doi:10.1109/icip.2016.7532690 · Cited by 75
McDonnell, M. D., Tissera, M. D., Vladusich, T., André, V. S., & Tapson, J. C. (2015) · PLoS ONE · doi:10.1371/journal.pone.0134254 · Cited by 89
“In order to loosely mimic this organisation of biological sensory systems, in this paper where we consider only image classification tasks, for each hidden unit we create randomly positioned and sized rectangular masks that are smaller than the overall image.”
BackgroundDeep Manifold Learning Combined With Convolutional Neural Networks for Action Recognition
Chen, X., Weng, J., Lu, W., Xu, J., & Weng, J. (2018) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2017.2740318 · Cited by 87
“All the subsampling layers apply max sampling as described in [57].”
MethodBilinear deep learning for image classification
Zhong, S.-H., Liu, Y., & Liu, Y. (2011) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/2072298.2072344 · Cited by 83
“We compare the performance of BDBN with other representative classifiers, including k-nearest neighbor (KNN), support vector machines (SVM) [46], transductive SVM (TSVM) [47], neural network (NN) [48], EmbedNN [28], Semi-DBN [45], DBN-rNCA [27], DDBN [29], and DCNN [37].”
MethodA Study of Complex Deep Learning Networks on High-Performance, Neuromorphic, and Quantum Computers
Potok, T. E., Schuman, C., Young, S., Patton, R., Spedalieri, F., Liu, J., Yao, K.-T., Rose, G., & Chakma, G. (2018) · ACM Journal on Emerging Technologies in Computing Systems · doi:10.1145/3178454 · Cited by 77
“Subsequent improvements to CNNs, such as dropout (Hinton et al. 2012), rectified linear units (Jarrett et al. 2009; Nair and Hinton 2010), and efficient GPU deep learning codes (Abadi et al. 2015; Al-Rfou et al. 2016; Jia et al. 2014), have allowed CNNs to achieve impressive results on a variety of…”
BackgroundHow agents see things: On visual representations in an emergent language game
Bouchacourt, D., & Baroni, M. (2018) · Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/d18-1119 · Cited by 78
“They are also somewhat correlated with the input, which we attribute to the fact that untrained neural networks can already extract relevant image features (Jarrett et al., 2009).”
BackgroundTextCaps: Handwritten Character Recognition With Very Small Datasets
Jayasundara, V., Jayasekara, S., Jayasekara, H., Rajasegaran, J., Seneviratne, S., & Rodrigo, R. (2019) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv.2019.00033 · Cited by 93
“Multiple works [7, 8, 9, 10, 11] have used CNN models on MNIST dataset and have achieved results in excess of 99% accuracy.”
MethodLearning to Read Chest X-Ray Images from 16000+ Examples Using CNN
Dong, Y., Pan, Y., Zhang, J., & Xu, W. (2017) · 2017 IEEE/ACM International Conference on Connected Health: Applications, Systems and Engineering Technologies (CHASE) · doi:10.1109/chase.2017.59 · Cited by 77
Robust 3D Hand Pose Estimation From Single Depth Images Using Multi-View CNNs
Ge, L., Liang, H., Yuan, J., & Thalmann, D. (2018) · IEEE Transactions on Image Processing · doi:10.1109/tip.2018.2834824 · Cited by 83
“The input projected images are first resized and padded to 96×96 pixels and filtered by local contrast normalization (LCN) [51] to normalize the contrast in the image.”
MethodOrimoloye, L. O., Sung, M., Ma, T., & JOHNSON, J. E. V. (2019) · Expert Systems with Applications · doi:10.1016/j.eswa.2019.112828 · Cited by 96
Deep Residual Network in Network
Alaeddine, H., & Malek, J. (2021) · Computational Intelligence and Neuroscience · doi:10.1155/2021/6659083 · Cited by 89
Karim, A., Lee, M., Balle, T., & Sattar, A. (2021) · Journal of Cheminformatics · doi:10.1186/s13321-021-00541-z · Cited by 99
“The number of units in each hidden layer is decreased gradually and a ReLu activation [46, 47] is applied at the end of each layer.”
Using grayscale images for object recognition with convolutional-recursive neural network
Bui, H. M., Lech, M., Cheng, E., Neville, K. L., & Burnett, L. (2016) · International Conference on Communications and Electronics (ICCE) · doi:10.1109/cce.2016.7562656 · Cited by 98
“proposed a high level of learning hierarchy to provide better features and improve recognition [14]; however, high computational complexity is required due to the many parameters that must be tuned.”
BackgroundNogueira, R. F., & Lotufo, R. (2014) · LA Referencia (Red Federada de Repositorios Institucionales de Publicaciones Científicas) · Cited by 78
An Overview of Convolutional Neural Network: Its Architecture and ApplicationsPreprint
Sakib, S., Ahmed, N., Kabir, A. J., & Ahmed, H. (2019) · Preprints.org · doi:10.20944/preprints201811.0546.v4 · Cited by 108
Flower classification via convolutional neural network
Liu, Y., Tang, F., Zhou, D., Meng, Y., & Dong, W. (2016) · International Symposium on Plant Growth Modeling, Simulation, Visualization and Applications (PMA) · doi:10.1109/fspma.2016.7818296 · Cited by 79
“Therefore, ConvNets have been substantially improving upon the state of the art in image classification and other recognition tasks [3, 4, 5,10, 20, 29].”
BackgroundYu, J., Weng, K., Liang, G., & Xie, G. (2013) · 2013 IEEE International Conference on Robotics and Biomimetics (ROBIO) · doi:10.1109/robio.2013.6739623 · Cited by 76
“Like almost every other neural networks CNNs are trained with a version of the back-propagation algorithm [13].”
MethodIterative deep learning for image set based face and object recognition
Shah, S. A. A., Bennamoun, M., & Boussaïd, F. (2015) · Neurocomputing · doi:10.1016/j.neucom.2015.10.004 · Cited by 75
Random curiosity-driven exploration in deep reinforcement learning
Li, J. K., Shi, X., Li, J., Zhang, X., & Wang, J. (2020) · Neurocomputing · doi:10.1016/j.neucom.2020.08.024 · Cited by 81
Automatic diet monitoring: a review of computer vision and wearable sensor-based methods
Hassannejad, H., Matrella, G., Ciampolini, P., De Munari, I., Mordonini, M., & Cagnoni, S. (2017) · International Journal of Food Sciences and Nutrition · doi:10.1080/09637486.2017.1283683 · Cited by 76
“In particular, convolutional neural networks (CNNs) (Jarrett et al. 2009) constitute one such class of models.”
BackgroundDeep Deconvolutional Networks for Scene Parsing
Mohan, R. (2014) · arXiv (Cornell University) · doi:10.48550/arxiv.1411.4101 · Cited by 53
Comparing state-of-the-art visual features on invariant object recognition tasks
Pinto, N., Barhomi, Y., Cox, D., & DiCarlo, J. J. (2011) · Proceedings · doi:10.1109/wacv.2011.5711540 · Cited by 86
“Caltech-101, PASCAL VOC) continue to be used almost exclusively as evidence of progress in solving object recognition [10, 9 , 22, 34].”
BackgroundInfluentialFast Deep Convolutional Face Detection in the Wild Exploiting Hard Sample Mining
Triantafyllidou, D., Nousi, P., & Tefas, A. (2017) · Big Data Research · doi:10.1016/j.bdr.2017.06.002 · Cited by 76
A fast face detection method via convolutional neural network
Guo, G., Wang, H., Yan, Y., Zheng, J. Y., & Li, B. H. (2019) · Neurocomputing · doi:10.1016/j.neucom.2018.02.110 · Cited by 81
In-Silico Molecular Binding Prediction for Human Drug Targets Using Deep Neural Multi-Task Learning
Lee, K., & Kim, D. (2019) · Genes · doi:10.3390/genes10110906 · Cited by 74
“Deep learning uses multiple layers with non-linear activation functions for effective learning of complex relationships hidden within the data [10].”
BackgroundImage Classification Using Convolutional Neural Networks With Multi-stage Feature
Yim, J., Ju, J., Jung, H., & Kim, J. (2015) · Advances in intelligent systems and computing · doi:10.1007/978-3-319-16841-8_52 · Cited by 82
Two-stage CNNs for computerized BI-RADS categorization in breast ultrasound images
Huang, Y., Han, L., Dou, H., Luo, H., Yuan, Z., Liu, Q., Zhang, J., & Yin, G. (2019) · BioMedical Engineering OnLine · doi:10.1186/s12938-019-0626-5 · Cited by 80
“And the extensively utilized activation methods in the CNN include the rectified linear unit (ReLU), sigmoid, and tanh [35].”
MethodHyperspectral Image Super-Resolution via Intrafusion Network
Hu, J., Jia, X., Li, Y., He, G., & Zhao, M. (2020) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2020.2982940 · Cited by 75
Heinlein, A., Klawonn, A., Lanser, M., & Weber, J. (2021) · GAMM-Mitteilungen · doi:10.1002/gamm.202100001 · Cited by 84
“Among the many different choices for the activation function 𝛼 , we choose the rectified linear unit [23, 34, 63] given by As input data for the neural network, we use samples, that is, point-wise evaluations of the coefficient function or the material distribution within the two subdomains…”
MethodNear-optimal interception strategy for orbital pursuit-evasion using deep reinforcement learning
Zhang, J., Zhang, K., Zhang, Y., Shi, H., Tang, L., & Li, M. (2022) · Acta Astronautica · doi:10.1016/j.actaastro.2022.05.057 · Cited by 73
When Ensemble Learning Meets Deep Learning: a New Deep Support Vector Machine for Classification
Qi, Z., Wang, B., Tian, Y., & Zhang, P. (2016) · Knowledge-Based Systems · doi:10.1016/j.knosys.2016.05.055 · Cited by 76
Neural networks designing neural networks: multi-objective hyper-parameter optimization
Smithson, S. C., Yang, G., Gross, W. J., & Meyer, B. H. (2016) · doi:10.1145/2966986.2967058 · Cited by 103
“Therefore, multiply-accumulate operations and memory accesses remain the dominant tasks in terms of cost [9, 12].”
BackgroundMultimodal learning using convolution neural network and Sparse Autoencoder
Vu, T. D., Yang, H.-J., Nguyen, V. Q., Oh, A., & Kim, M.-S. (2017) · IEEE International Conference on Big Data and Smart Computing (BigComp) · doi:10.1109/bigcomp.2017.7881683 · Cited by 90
“The cost function is minimized using gradient descent with mini batches [10] and the algorithm is expected to converge faster than with full batches.”
BackgroundMethodInfluentialPower Line Recognition From Aerial Images With Deep Learning
Yetgin, Ö. E., Benlıgıray, B., & Gerek, Ö. N. (2018) · IEEE Transactions on Aerospace and Electronic Systems · doi:10.1109/taes.2018.2883879 · Cited by 77
“Then, it is reasonable to expect untrained nets to perform reasonably well in power line recognition, as they do in other tasks [49].”
BackgroundMonarch Butterfly Optimization Based Convolutional Neural Network Design
Bačanin, N., Bezdan, T., Tuba, E., Strumberger, I., & Tuba, M. (2020) · Mathematics · doi:10.3390/math8060936 · Cited by 77
“52 CNN (3 conv maxout, 1 dense) with dropout [101] 0.”
One-dimensional convolutional neural network-based damage detection in structural joints
Sharma, S., & Sen, S. (2020) · Journal of Civil Structural Health Monitoring · doi:10.1007/s13349-020-00434-z · Cited by 78
“It has been employed for various vision applications [21, 30, 31] to demonstrate that CNNs can easily identify unique attributes even from complex and uncorrelated signals which can never be achieved by even a human expert.”
BackgroundFarabet, C., Paz, R., Pérez‐Carrasco, J. A., Zamarreño-Ramos, C., Linares-Barranco, A., LeCun, Y., Culurciello, E., Serrano‐Gotarredona, T., & Linares-Barranco, B. (2012) · Frontiers in Neuroscience · doi:10.3389/fnins.2012.00032 · Cited by 74
“ConvNets have been successfully used in many recognition and classification tasks including document recognition (LeCun et al., 1998a),objectrecognition(HuangandLeCun,2006;Ranzatoetal., 2007;Jarrettetal.,2009),facedetection(Osadchyetal.,2005),and robotnavigation(Hadselletal.,2007,2009).”
MethodSelf-citationExtreme learning machine with kernel model based on deep learning
Ding, S., Guo, L., & Hou, Y. (2016) · Neural Computing and Applications · doi:10.1007/s00521-015-2170-y · Cited by 74
“Research has shown that the effect of network is mainly determined by its structure [12].”
Background
Metadata
- Authors
- Jarrett, K., Kavukcuoglu, K., Ranzato, M. A., & LeCun, Y.
- Affiliations
- Courant Institute of Mathematical Sciences (US)
- ORCID
- LeCun, Yann
- Venue
- 2009 IEEE 12th International Conference on Computer Vision
- Publisher
- IEEE
- Published
- 2009-09
- Pages
- 2146-2153
- OpenAlex
- W2546302380
- Semantic Scholar
- 1f88427d7aa8225e47f946ac41a0667d7b69ac52
- Cited by
- 2,153
- References
- 57
- Citations by year
- 2026: 20 · 2025: 48 · 2024: 82 · 2023: 114 · 2022: 124 · 2021: 194 · 2020: 207 · 2019: 205 · 2018: 211 · 2017: 210 · 2016: 198 · 2015: 162 · 2014: 137 · 2013: 86 · 2012: 69
- FWCI
- 42.18
- Type
- conference paper
- Language
- en
- Open access
- green
- hal.science
- Licence
- cc-by-nc
- Locations
- IEEE/CVF International Conference on Computer Vision (ICCV) (publishedVersion)
- HAL (submittedVersion, cc-by-nc)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Image and Video Retrieval Techniques
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Image Retrieval and Classification Techniques
- Keywords
- object recognition, filter banks, rectification, feature pooling, unsupervised pre-training, Caltech-101, MNIST, image classification
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.