Simulating a Primary Visual Cortex at the Front of CNNs Improves Robustness to Image Perturbations
Open accessThe DOI matches its Crossref record; 33 of the 33 references it cites were found in a registry or an index.
Abstract
Current state-of-the-art object recognition models are largely based on convolutional neural network (CNN) architectures, which are loosely inspired by the primate visual system. However, these CNNs can be fooled by imperceptibly small, explicitly crafted perturbations, and struggle to recognize objects in corrupted images that are easily recognized by humans. Here, by making comparisons with primate neural data, we first observed that CNN models with a neural hidden layer that better matches primate primary visual cortex (V1) are also more robust to adversarial attacks. Inspired by this observation, we developed VOneNets, a new class of hybrid CNN vision models. Each VOneNet contains a fixed weight neural network front-end that simulates primate V1, called the VOneBlock, followed by a neural network back-end adapted from current CNN vision models. The VOneBlock is based on a classical neuroscientific model of V1: the linear-nonlinear-Poisson model, consisting of a biologically-constrained Gabor filter bank, simple and complex cell nonlinearities, and a V1 neuronal stochasticity generator. After training, VOneNets retain high ImageNet performance, but each is substantially more robust, outperforming the base CNNs and state-of-the-art methods by 18% and 3%, respectively, on a conglomerate benchmark of perturbations comprised of white box adversarial attacks and common image corruptions. Finally, we show that all components of the VOneBlock work in synergy to improve robustness. While current CNN architectures are arguably brain-inspired, the results presented here demonstrate that more precisely mimicking just one stage of the primate visual system leads to new gains in ImageNet-level computer vision applications.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
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
ImageNet 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
LeCun, Y., Bengio, Y., & Hinton, G. (2015) · Nature · doi:10.1038/nature14539 · Cited by 84,126
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
Towards Evaluating the Robustness of Neural Networks
Carlini, N., & Wagner, D. (2017) · 2017 IEEE Symposium on Security and Privacy (SP) · doi:10.1109/sp.2017.49 · Cited by 8,859
Intriguing properties of neural networks
Szegedy, C., Zaremba, W., Sutskever, I., Bruna, J., Erhan, D., Goodfellow, I., & Fergus, R. (2013) · arXiv · doi:10.48550/arxiv.1312.6199 · Cited by 8,143
ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design
Ma, N., Zhang, X., Zheng, H.-T., & Sun, J. (2018) · Lecture notes in computer science · doi:10.1007/978-3-030-01264-9_8 · Cited by 7,356 · cited 1×
“…networks with standard ImageNet training [70] including AlexNet [1], VGG [3], ResNet [4], ResNeXt [71], DenseNet [72], SqueezeNet [73], ShuffleNet [74], and MnasNet [75], as well as several ResNet50 models with specialized training routines, such as adversarial training with L ∞ ( (cid:107) δ ∞ =…”
MethodTowards Deep Learning Models Resistant to Adversarial Attacks
Madry, A., Makelov, A., Schmidt, L., Tsipras, D., & Vladu, A. (2017) · arXiv · doi:10.48550/arxiv.1706.06083 · Cited by 6,602
SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5MB model size
Iandola, F. N., Han, S., Moskewicz, M. W., Ashraf, K., Dally, W. J., & Keutzer, K. (2016) · arXiv · doi:10.48550/arxiv.1602.07360 · Cited by 6,201
Adversarial examples in the physical world
Kurakin, A., Goodfellow, I. J., & Bengio, S. (2018) · Artificial Intelligence Safety and Security · doi:10.1201/9781351251389-8 · Cited by 5,375
DeepLabCut: markerless pose estimation of user-defined body parts with deep learning
Mathis, A., Mamidanna, P., Cury, K. M., Abe, T., Murthy, V. N., Mathis, M. W., & Bethge, M. (2018) · Nature Neuroscience · doi:10.1038/s41593-018-0209-y · Cited by 5,187
Spatiotemporal energy models for the perception of motion
Adelson, E. H., & Bergen, J. R. (1985) · Journal of the Optical Society of America A · doi:10.1364/josaa.2.000284 · Cited by 3,442
MnasNet: Platform-Aware Neural Architecture Search for Mobile
Tan, M., Chen, B., Pang, R., Vasudevan, V., Sandler, M., Howard, A., & Le, Q. V. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.00293 · Cited by 3,416 · cited 1×
“…ImageNet training [70] including AlexNet [1], VGG [3], ResNet [4], ResNeXt [71], DenseNet [72], SqueezeNet [73], ShuffleNet [74], and MnasNet [75], as well as several ResNet50 models with specialized training routines, such as adversarial training with L ∞ ( (cid:107) δ ∞ = 4 / 255 and…”
MethodRepresentational similarity analysis-connecting the branches of systems neuroscience
Kriegeskorte, N. (2008) · Frontiers in Systems Neuroscience · doi:10.3389/neuro.06.004.2008 · Cited by 3,355
Hierarchical models of object recognition in cortex
Riesenhuber, M., & Poggio, T. (1999) · Nature Neuroscience · doi:10.1038/14819 · Cited by 3,309 · cited 1×
“Early hierarchical models of object recognition, which incorporated these type of V1 models as their early layers, were used with some success in modeling neuronal responses in the ventral stream and object recognition behavior [50, 51].”
MethodObfuscated gradients give a false sense of security: Circumventing defenses to adversarial examples
Athalye, A., Carlini, N., & Wagner, D. (2018) · arXiv · doi:10.48550/arxiv.1802.00420 · Cited by 2,172 · cited 6×
“[68] and Carlini et. al.”
BackgroundMethodInfluentialPerformance-optimized hierarchical models predict neural responses in higher visual cortex
Yamins, D. L. K., Hong, H., Cadieu, C. F., Solomon, E. A., Seibert, D., & DiCarlo, J. J. (2014) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1403112111 · Cited by 1,924
Self-citationSoftky, W., & Koch, C. (1993) · The Journal of Neuroscience · doi:10.1523/jneurosci.13-01-00334.1993 · Cited by 1,689 · cited 2×
“In awake monkeys, the mean spike count (averaged over many repetitions) depends on the presented image, and the spike train for each trial is approximately Poisson: the spike count variance is equal to the mean [82].”
BackgroundJones, J. P., & Palmer, L. A. (1987) · Journal of Neurophysiology · doi:10.1152/jn.1987.58.6.1233 · Cited by 1,658 · cited 1×
“Early approaches consisting of hand-designed Gabor filters were successful in predicting V1 simple cell [46] and complex cell [47] responses to relatively simple stimuli.”
MethodUsing goal-driven deep learning models to understand sensory cortex
Yamins, D. L. K., & DiCarlo, J. J. (2016) · Nature Neuroscience · doi:10.1038/nn.4244 · Cited by 1,621
Self-citationNeuroscience-Inspired Artificial Intelligence
Hassabis, D., Kumaran, D., Summerfield, C., & Botvinick, M. (2017) · Neuron · doi:10.1016/j.neuron.2017.06.011 · Cited by 1,557 · cited 2×
“How can we develop CNNs that robustly generalize like human vision? Incorporating biological constraints into CNNs to make them behave more in line with primate vision is an active field of research [33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43].”
BackgroundSpatial frequency selectivity of cells in macaque visual cortex
De Valois, R. L., Albrecht, D. G., & Thorell, L. G. (1982) · Vision Research · doi:10.1016/0042-6989(82)90113-4 · Cited by 1,489
Benchmarking Neural Network Robustness to Common Corruptions and Perturbations
Hendrycks, D., & Dietterich, T. (2019) · arXiv · doi:10.48550/arxiv.1903.12261 · Cited by 1,335
Deep supervised, but not unsupervised, models may explain IT cortical representation
Khaligh-Razavi, S.-M., & Kriegeskorte, N. (2014) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1003915 · Cited by 1,204
Highly selective receptive fields in mouse visual cortex
Niell, C. M., & Stryker, M. P. (2008) · The Journal of Neuroscience · doi:10.1523/jneurosci.0623-08.2008 · Cited by 1,081 · cited 1×
“Visual acuity in mice is much lower than in macaques and humans [87, 79], suggesting that vision in the mouse may serve different behavioral functions than in primates.”
BackgroundGuclu, U., & van Gerven, M. A. J. (2015) · Journal of Neuroscience · doi:10.1523/jneurosci.5023-14.2015 · Cited by 961
Feature denoising for improving adversarial robustness
Xie, C., Wu, Y., Maaten, L. V. D., Yuille, A. L., & He, K. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.00059 · Cited by 940
Cadieu, C. F., Hong, H., Yamins, D. L. K., Pinto, N., Ardila, D., Solomon, E. A., Majaj, N. J., & DiCarlo, J. J. (2014) · PLoS Computational Biology · doi:10.1371/journal.pcbi.1003963 · Cited by 824
Self-citationHow does nature program neuron types?
Borst, A. (2008) · Frontiers in Neuroscience · doi:10.3389/neuro.01.016.2008 · Cited by 705
Adversarial Robustness Toolbox v1.0.0
Nicolae, M.-I., Sinn, M., Tran, M. N., Buesser, B., Rawat, A., Wistuba, M., Zantedeschi, V., Baracaldo, N., Chen, B., Ludwig, H., et al. (2018) · arXiv · doi:10.48550/arxiv.1807.01069 · Cited by 594
PixelDefend: Leveraging Generative Models to Understand and Defend against Adversarial Examples
Song, Y., Kim, T., Nowozin, S., Ermon, S., & Kushman, N. (2017) · arXiv · doi:10.48550/arxiv.1710.10766 · Cited by 562
On Adaptive Attacks to Adversarial Example Defenses
Tramer, F., Carlini, N., Brendel, W., & Madry, A. (2020) · arXiv · doi:10.48550/arxiv.2002.08347 · Cited by 486
Benchmarking Robustness in Object Detection: Autonomous Driving when Winter is Coming
Michaelis, C., Mitzkus, B., Geirhos, R., Rusak, E., Bringmann, O., Ecker, A. S., Bethge, M., & Brendel, W. (2019) · arXiv · doi:10.48550/arxiv.1907.07484 · Cited by 324
Cited by
The neuroconnectionist research programme
Doerig, A., Sommers, R. P., Seeliger, K., Richards, B., Ismael, J., Lindsay, G. W., Kording, K. P., Konkle, T., van Gerven, M. A. J., Kriegeskorte, N., et al. (2023) · Nature Reviews Neuroscience · doi:10.1038/s41583-023-00705-w · Cited by 233
Deep problems with neural network models of human vision
Bowers, J. S., Malhotra, G., Dujmović, M., Llera Montero, M., Tsvetkov, C., Biscione, V., Puebla, G., Adolfi, F., Hummel, J. E., Heaton, R. F., et al. (2023) · Behavioral and Brain Sciences · doi:10.1017/s0140525x22002813 · Cited by 204
Performance vs. competence in human–machine comparisons
Firestone, C. (2020) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1905334117 · Cited by 150
Partial success in closing the gap between human and machine vision
Geirhos, R., Narayanappa, K., Mitzkus, B., Thieringer, T., Bethge, M., Wichmann, F. A., & Brendel, W. (2021) · arXiv · doi:10.48550/arxiv.2106.07411 · Cited by 129
Toward Transparent AI: A Survey on Interpreting the Inner Structures of Deep Neural Networks
Räuker, T., Ho, A., Casper, S., & Hadfield-Menell, D. (2023) · 2023 IEEE Conference on Secure and Trustworthy Machine Learning (SaTML) · doi:10.1109/satml54575.2023.00039 · Cited by 122
Metadata
- Authors
- Dapello, J., Marques, T., Schrimpf, M., Geiger, F., Cox, D. D., & DiCarlo, J. J.
- Venue
- bioRxiv (Cold Spring Harbor Laboratory)
- Publisher
- openRxiv
- Published
- 2020-06-17
- OpenAlex
- W3035475567
- Cited by
- 117
- References
- 33
- Type
- posted content
- Language
- en
- Open access
- green
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.