We propose RepMLP, a multi-layer-perceptron-style neural network building block for image recognition, which is composed of a series of fully-connected (FC) layers. Compared to convolutional layers, FC layers are more efficient, better at modeling the long-range dependencies and positional patterns, but worse at capturing the local structures, hence usually less favored for image recognition. We propose a structural re-parameterization technique that adds local prior into an FC to make it powerful for image recognition. Specifically, we construct convolutional layers inside a RepMLP during training and merge them into the FC for inference. On CIFAR, a simple pure-MLP model shows performance very close to CNN. By inserting RepMLP in traditional CNN, we improve ResNets by 1.8% accuracy on ImageNet, 2.9% for face recognition, and 2.3% mIoU on Cityscapes with lower FLOPs. Our intriguing findings highlight that combining the global representational capacity and positional perception of FC with the local prior of convolution can improve the performance of neural network with faster speed on both the tasks with translation invariance (e.g., semantic segmentation) and those with aligned images and positional patterns (e.g., face recognition). The code and models are available at //github.com/DingXiaoH/RepMLP.
相關內容
Visual recognition has been dominated by convolutionalneural networks (CNNs) for years. Though recently the pre-vailing vision transformers (ViTs) have shown great poten-tial of self-attention based models in ImageNet classifica-tion, their performance is still inferior to latest SOTA CNNsif no extra data are provided. In this work, we aim to closethe performance gap and demonstrate that attention-basedmodels are indeed able to outperform CNNs. We found thatthe main factor limiting the performance of ViTs for Ima-geNet classification is their low efficacy in encoding fine-level features into the token representations. To resolvethis, we introduce a noveloutlook attentionand present asimple and general architecture, termed Vision Outlooker(VOLO). Unlike self-attention that focuses on global depen-dency modeling at a coarse level, the outlook attention aimsto efficiently encode finer-level features and contexts intotokens, which are shown to be critical for recognition per-formance but largely ignored by the self-attention. Experi-ments show that our VOLO achieves 87.1% top-1 accuracyon ImageNet-1K classification, being the first model exceed-ing 87% accuracy on this competitive benchmark, withoutusing any extra training data. In addition, the pre-trainedVOLO transfers well to downstream tasks, such as seman-tic segmentation. We achieve 84.3% mIoU score on thecityscapes validation set and 54.3% on the ADE20K valida-tion set. Code is available at //github.com/sail-sg/volo.
In this paper, we present Vision Permutator, a conceptually simple and data efficient MLP-like architecture for visual recognition. By realizing the importance of the positional information carried by 2D feature representations, unlike recent MLP-like models that encode the spatial information along the flattened spatial dimensions, Vision Permutator separately encodes the feature representations along the height and width dimensions with linear projections. This allows Vision Permutator to capture long-range dependencies along one spatial direction and meanwhile preserve precise positional information along the other direction. The resulting position-sensitive outputs are then aggregated in a mutually complementing manner to form expressive representations of the objects of interest. We show that our Vision Permutators are formidable competitors to convolutional neural networks (CNNs) and vision transformers. Without the dependence on spatial convolutions or attention mechanisms, Vision Permutator achieves 81.5% top-1 accuracy on ImageNet without extra large-scale training data (e.g., ImageNet-22k) using only 25M learnable parameters, which is much better than most CNNs and vision transformers under the same model size constraint. When scaling up to 88M, it attains 83.2% top-1 accuracy. We hope this work could encourage research on rethinking the way of encoding spatial information and facilitate the development of MLP-like models. Code is available at //github.com/Andrew-Qibin/VisionPermutator.
Neuromorphic neural network processors, in the form of compute-in-memory crossbar arrays of memristors, or in the form of subthreshold analog and mixed-signal ASICs, promise enormous advantages in compute density and energy efficiency for NN-based ML tasks. However, these technologies are prone to computational non-idealities, due to process variation and intrinsic device physics. This degrades the task performance of networks deployed to the processor, by introducing parameter noise into the deployed model. While it is possible to calibrate each device, or train networks individually for each processor, these approaches are expensive and impractical for commercial deployment. Alternative methods are therefore needed to train networks that are inherently robust against parameter variation, as a consequence of network architecture and parameters. We present a new adversarial network optimisation algorithm that attacks network parameters during training, and promotes robust performance during inference in the face of parameter variation. Our approach introduces a regularization term penalising the susceptibility of a network to weight perturbation. We compare against previous approaches for producing parameter insensitivity such as dropout, weight smoothing and introducing parameter noise during training. We show that our approach produces models that are more robust to targeted parameter variation, and equally robust to random parameter variation. Our approach finds minima in flatter locations in the weight-loss landscape compared with other approaches, highlighting that the networks found by our technique are less sensitive to parameter perturbation. Our work provides an approach to deploy neural network architectures to inference devices that suffer from computational non-idealities, with minimal loss of performance. ...
Multi-label image classification (MLIC) is a fundamental and practical task, which aims to assign multiple possible labels to an image. In recent years, many deep convolutional neural network (CNN) based approaches have been proposed which model label correlations to discover semantics of labels and learn semantic representations of images. This paper advances this research direction by improving both the modeling of label correlations and the learning of semantic representations. On the one hand, besides the local semantics of each label, we propose to further explore global semantics shared by multiple labels. On the other hand, existing approaches mainly learn the semantic representations at the last convolutional layer of a CNN. But it has been noted that the image representations of different layers of CNN capture different levels or scales of features and have different discriminative abilities. We thus propose to learn semantic representations at multiple convolutional layers. To this end, this paper designs a Multi-layered Semantic Representation Network (MSRN) which discovers both local and global semantics of labels through modeling label correlations and utilizes the label semantics to guide the semantic representations learning at multiple layers through an attention mechanism. Extensive experiments on four benchmark datasets including VOC 2007, COCO, NUS-WIDE, and Apparel show a competitive performance of the proposed MSRN against state-of-the-art models.
Traditional image resizing methods usually work in pixel space and use various saliency measures. The challenge is to adjust the image shape while trying to preserve important content. In this paper we perform image resizing in feature space where the deep layers of a neural network contain rich important semantic information. We directly adjust the image feature maps, extracted from a pre-trained classification network, and reconstruct the resized image using a neural-network based optimization. This novel approach leverages the hierarchical encoding of the network, and in particular, the high-level discriminative power of its deeper layers, that recognizes semantic objects and regions and allows maintaining their aspect ratio. Our use of reconstruction from deep features diminishes the artifacts introduced by image-space resizing operators. We evaluate our method on benchmarks, compare to alternative approaches, and demonstrate its strength on challenging images.
Deep learning has shown promising results in medical image analysis, however, the lack of very large annotated datasets confines its full potential. Although transfer learning with ImageNet pre-trained classification models can alleviate the problem, constrained image sizes and model complexities can lead to unnecessary increase in computational cost and decrease in performance. As many common morphological features are usually shared by different classification tasks of an organ, it is greatly beneficial if we can extract such features to improve classification with limited samples. Therefore, inspired by the idea of curriculum learning, we propose a strategy for building medical image classifiers using features from segmentation networks. By using a segmentation network pre-trained on similar data as the classification task, the machine can first learn the simpler shape and structural concepts before tackling the actual classification problem which usually involves more complicated concepts. Using our proposed framework on a 3D three-class brain tumor type classification problem, we achieved 82% accuracy on 191 testing samples with 91 training samples. When applying to a 2D nine-class cardiac semantic level classification problem, we achieved 86% accuracy on 263 testing samples with 108 training samples. Comparisons with ImageNet pre-trained classifiers and classifiers trained from scratch are presented.
In this paper, we propose a novel fully convolutional two-stream fusion network (FCTSFN) for interactive image segmentation. The proposed network includes two sub-networks: a two-stream late fusion network (TSLFN) that predicts the foreground at a reduced resolution, and a multi-scale refining network (MSRN) that refines the foreground at full resolution. The TSLFN includes two distinct deep streams followed by a fusion network. The intuition is that, since user interactions are more direction information on foreground/background than the image itself, the two-stream structure of the TSLFN reduces the number of layers between the pure user interaction features and the network output, allowing the user interactions to have a more direct impact on the segmentation result. The MSRN fuses the features from different layers of TSLFN with different scales, in order to seek the local to global information on the foreground to refine the segmentation result at full resolution. We conduct comprehensive experiments on four benchmark datasets. The results show that the proposed network achieves competitive performance compared to current state-of-the-art interactive image segmentation methods.
In this work, we evaluate the use of superpixel pooling layers in deep network architectures for semantic segmentation. Superpixel pooling is a flexible and efficient replacement for other pooling strategies that incorporates spatial prior information. We propose a simple and efficient GPU-implementation of the layer and explore several designs for the integration of the layer into existing network architectures. We provide experimental results on the IBSR and Cityscapes dataset, demonstrating that superpixel pooling can be leveraged to consistently increase network accuracy with minimal computational overhead. Source code is available at //github.com/bermanmaxim/superpixPool
For the challenging semantic image segmentation task the most efficient models have traditionally combined the structured modelling capabilities of Conditional Random Fields (CRFs) with the feature extraction power of CNNs. In more recent works however, CRF post-processing has fallen out of favour. We argue that this is mainly due to the slow training and inference speeds of CRFs, as well as the difficulty of learning the internal CRF parameters. To overcome both issues we propose to add the assumption of conditional independence to the framework of fully-connected CRFs. This allows us to reformulate the inference in terms of convolutions, which can be implemented highly efficiently on GPUs. Doing so speeds up inference and training by a factor of more then 100. All parameters of the convolutional CRFs can easily be optimized using backpropagation. To facilitating further CRF research we make our implementation publicly available. Please visit: //github.com/MarvinTeichmann/ConvCRF
Spatial pyramid pooling module or encode-decoder structure are used in deep neural networks for semantic segmentation task. The former networks are able to encode multi-scale contextual information by probing the incoming features with filters or pooling operations at multiple rates and multiple effective fields-of-view, while the latter networks can capture sharper object boundaries by gradually recovering the spatial information. In this work, we propose to combine the advantages from both methods. Specifically, our proposed model, DeepLabv3+, extends DeepLabv3 by adding a simple yet effective decoder module to refine the segmentation results especially along object boundaries. We further explore the Xception model and apply the depthwise separable convolution to both Atrous Spatial Pyramid Pooling and decoder modules, resulting in a faster and stronger encoder-decoder network. We demonstrate the effectiveness of the proposed model on the PASCAL VOC 2012 semantic image segmentation dataset and achieve a performance of 89% on the test set without any post-processing. Our paper is accompanied with a publicly available reference implementation of the proposed models in Tensorflow.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191