Systematic evaluation of CNN advances on the ImageNet.

The paper systematically studies the impact of a range of recent advances in CNN architectures and learning methods on the object categorization (ILSVRC) problem. ... The evalution tests the influence of the following choices of the architecture: non-linearity (ReLU, ELU, maxout, compatibility with batch normalization), pooling variants (stochastic, max, average, mixed ... Acknowledgements The authors were supported by The Czech Science Foundation Project GACR P103/12/G084 and CTU student grant SGS15/155/OHK3/2T/13. ...

doi:10.1016/j.cviu.2017.05.007 fatcat:4u2un2jkp5aapc5wfoxihjxtey

Multiple Versions

The experimental results on ImageNet and six robustness benchmarks show the advanced robustness and generalization ability of RVT compared with previous ViTs and state-of-the-art CNNs. ... In this work, we conduct systematic evaluation on components of ViTs in terms of their impact on robustness to adversarial examples, common corruptions and distribution shifts. ... corruptions with five levels of severity. 3) for out-of-distribution robustness, we evaluate on ImageNet-R [14] and ImageNet-Sketch [15] . ...

arXiv:2105.07926v3 fatcat:xocuup3wqvae7o4n5npuspmbcu

Multiple Versions

We use two popular CNNs for our experiments -- Imagenet CNN and a modified version of LeNet CNN. ... We evaluate our recognition framework on a publicly available benchmark database containing thousands of freehand sketches depicting everyday objects. ... The sketch features are obtained by tapping the output of the inner-product layer of the trained Lenet CNN with the sketch as the input as in the case of Imagenet CNN. ...

arXiv:1502.00254v2 fatcat:h6jegxkqbrgi5pewdtvm7xothq

Multiple Versions

We then demonstrate that the same standard architecture (ResNet-50) that learns a texture-based representation on ImageNet is able to learn a shape-based representation instead when trained on "Stylized-ImageNet ... We here put these conflicting hypotheses to a quantitative test by evaluating CNNs and human observers on images with a texture-shape cue conflict. ... We would like to thank Dan Hendrycks for providing the results of Table 4 (corruption robustness of our models on ImageNet-C). ...

arXiv:1811.12231v2 fatcat:63xbjleue5gmbattknf5skegum

Multiple Versions

We found that while the best algorithms explain ~70% of the variance in the perceptual data, all the algorithms we tested make systematic errors on several types of objects. ... Here, we demonstrate two key conceptual advances: First, we show that most machine vision models are systematically different from human object perception. ... ACKNOWLEDGMENTS We thank Krithika Mohan and N Apurva Ratan Murty for sharing their data for inclusion in the dataset, and members of the Vision Lab for insightful discussions. ...

doi:10.1109/tpami.2020.3008107 pmid:32750809 pmcid:PMC7611439 fatcat:apeyzx2dfba4pblnil4g5rd75e

The image captioning task is increasingly prevalent in artificial intelligence applications for medicine. One important application is clinical report generation from chest radiographs. ... In this paper we demonstrate the importance of domain specific pre-training and propose a modified transformer architecture for the medical image captioning task. ... Here we propose a systematic investigation of model performance, especially related to its clinical success on each of the 14 radiological observations, by whether its feature extractor is pretrained on ...

arXiv:2205.02841v1 fatcat:fltlbdqqebb4vmcin7yccbnifi

Open Access

Deep learning techniques, such as convolutional neural networks (CNNs), generative adversarial networks (GANs), and graph neural networks (GNNs) have, over the past decade, changed the accuracy of prediction ... In recent years, the application of deep learning techniques in computer vision tasks in pathology has demonstrated extraordinary potential in assisting clinicians, automating diagnoses, and reducing costs ... Acknowledgments: We would like to thank all members of the Fenyö laboratory and the administration team of ISG at NYU. Conflicts of Interest: The authors declare no conflict of interest. ...

doi:10.3390/biomedinformatics2010010 fatcat:bagnt7eqgnelhnkj5dq7i4bjdu

Open Access

Adversarial examples are inevitable on the road of pervasive applications of deep neural networks (DNN). ... Lastly, we introduce recent arguments on potential costs of adversarial training which have attracted wide attention from the research community. ... evaluated the recent state-of-art ImageNet-based DNN models on multiple robustness metrics. ...

arXiv:2011.01539v1 fatcat:e3o47epftbc2rebpdx5yotzriy

In this work, we propose a comprehensive study that systematically evaluates FVs and CNNs for image retrieval. ... For CNNs, we focus on understanding the impact of depth, architecture and training data on retrieval results. ... CONCLUSIONS In this work, we proposed a systematic and in-depth evaluation of FV and CNN pipelines for image retrieval. ...

arXiv:1508.02496v3 fatcat:rferjnfn3vdh3ki3ew2gw3qj3e

Multiple Versions

The ImageNet pre-training initialization is the de-facto standard for object detection. ... Experiment results show that direct pre-training accelerates the pre-training phase by more than 11x on COCO dataset while with even +1.8mAP compared to ImageNet pre-training. ... In this paper, if without specification, we evaluate our approach on MS-COCO [8] datasets. We train the models on the COCO train2017 split, and evaluate on COCO val2017 split. ...

arXiv:2106.03112v1 fatcat:tigh77gq2jbovjbci2r5cgbi6e

RepLKNet also shows nice scalability to big data and large models, obtaining 87.8% top-1 accuracy on ImageNet and 56.0% mIoU on ADE20K, which is very competitive among the state-of-the-arts with similar ... RepLKNet greatly closes the performance gap between CNNs and ViTs, e.g., achieving comparable or superior results than Swin Transformer on ImageNet and a few typical downstream tasks, with lower latency ... To answer this question, we systematically explore the large kernel design of CNNs. ...

arXiv:2203.06717v4 fatcat:n4fz654m4zdw5hhgr3xyb745ii

Multiple Versions

Then on each task we compare the performance of two neural networks, one with top-down attention and one without. ... We make publicly available our code and results, along with statistics relevant to ImageNet-based experiments beyond this one. ... Evaluation: comparing the accuracy of two CNNs, one with top-down attention and one without. 5. Analysis: linear regression with the task-set properties as covariates and accuracy as response. ...

arXiv:2106.11339v1 fatcat:xswj7d66uvgatkg5x6kcoitvhu

Open Access

To this end, we introduce ImageNet-Instagram, a filtered version of ImageNet, where 20 popular Instagram filters are applied to each image in ImageNet. ... We further demonstrate the module can be readily plugged into modern CNN architectures together with skip connections. ... To evaluate the performance of modern CNN architectures on these filtered images, we run a ResNet50 (He et al. 2016 ) pretrained from IMAGENET on the validation set of IMAGENET-INSTAGRAM directly. ...

doi:10.1609/aaai.v34i07.6928 fatcat:bielcqrqxrgffk3uh5c3wqjcye

To this end, we introduce ImageNet-Instagram, a filtered version of ImageNet, where 20 popular Instagram filters are applied to each image in ImageNet. ... We further demonstrate the module can be readily plugged into modern CNN architectures together with skip connections. ... Conclusion We presented a study on how popular filters that are prevalent on social media affect the performance of pretrained modern CNN models. ...

arXiv:1912.13000v1 fatcat:623kzvc4mrc7llp3v4y46emdy4

of CNNs. ... Many efforts have been recently devoted to complementing CNNs with the global modeling ability, especially by a family of works on global feature interaction. ... We follow [2] to evaluate models trained on ImageNet on the overlapped classes. Performance Results. ...

arXiv:1910.05577v4 fatcat:6igeeru6yzcg3fgm444klrpdza

Multiple Versions

Systematic evaluation of convolution neural network advances on the Imagenet

Preserved Fulltext

Other Versions

Towards Robust Vision Transformer [article]

Preserved Fulltext

Other Versions

Freehand Sketch Recognition Using Deep Features [article]

Preserved Fulltext

Other Versions

ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness [article]

Preserved Fulltext

Other Versions

Improving Machine Vision using Human Perceptual Representations: The Case of Planar Reflection Symmetry for Object Classification

Preserved Fulltext

Understanding Transfer Learning for Chest Radiograph Clinical Report Generation with Modified Transformer Architectures [article]

Preserved Fulltext

Deep Learning and Its Applications in Computational Pathology

Preserved Fulltext

Recent Advances in Understanding Adversarial Robustness of Deep Neural Networks [article]

Preserved Fulltext

A Practical Guide to CNNs and Fisher Vectors for Image Instance Retrieval [article]

Preserved Fulltext

Other Versions

Rethinking Training from Scratch for Object Detection [article]

Preserved Fulltext

Scaling Up Your Kernels to 31x31: Revisiting Large Kernel Design in CNNs [article]

Preserved Fulltext

Other Versions

Understanding top-down attention using task-oriented ablation design [article]

Preserved Fulltext

Recognizing Instagram Filtered Images with Feature De-Stylization

Preserved Fulltext

Recognizing Instagram Filtered Images with Feature De-stylization [article]

Preserved Fulltext

Context-Gated Convolution [article]

Preserved Fulltext

Other Versions