The problem of robustness under location deformations for deep convolutional neural networks (DCNNs) is of great theoretical and practical interest. This issue has been studied in pioneering works, especially for scattering-type architectures, for deformation vector fields $\tau(x)$ with some regularity - at least $C^1$. Here we address this issue for any field $\tau\in L^\infty(\mathbb{R}^d;\mathbb{R}^d)$, without any additional regularity assumption, hence including the case of wild irregular deformations such as a noise on the pixel location of an image. We prove that for signals in multiresolution approximation spaces $U_s$ at scale $s$, whenever the network is Lipschitz continuous (regardless of its architecture), stability in $L^2$ holds in the regime $\|\tau\|_{L^\infty}/s\ll 1$, essentially as a consequence of the uncertainty principle. When $\|\tau\|_{L^\infty}/s\gg 1$ instability can occur even for well-structured DCNNs such as the wavelet scattering networks, and we provide a sharp upper bound for the asymptotic growth rate. The stability results are then extended to signals in the Besov space $B^{d/2}_{2,1}$ tailored to the given multiresolution approximation. We also consider the case of more general time-frequency deformations. Finally, we provide stochastic versions of the aforementioned results, namely we study the issue of stability in mean when $\tau(x)$ is modeled as a random field (not bounded, in general) with with identically distributed variables $|\tau(x)|$, $x\in\mathbb{R}^d$.
Borneo indigenous communities are blessed with rich craft heritage. One such examples is the Iban's plaited mat craft. There have been many efforts by UNESCO and the Sarawak Government to preserve and promote the craft. One such method is by developing a mobile app capable of recognising the different mat motifs. As a first step towards this aim, we presents a novel image dataset consisting of seven mat motif classes. Each class possesses a unique variation of chevrons, diagonal shapes, symmetrical, repetitive, geometric and non geometric patterns. In this study, the performance of the Scale invariant feature transform (SIFT) descriptor is evaluated against five common image deformations, i.e., zoom and rotation, viewpoint, image blur, JPEG compression and illumination. Using our dataset, SIFT performed favourably with test sequences belonging to Illumination changes, Viewpoint changes, JPEG compression and Zoom and Rotation. However, it did not performed well with Image blur test sequences with an average of 1.61 percents retained pairwise matching after blurring with a Gaussian kernel of 8.0 radius.
Backdoor attacks have been considered a severe security threat to deep learning. Such attacks can make models perform abnormally on inputs with predefined triggers and still retain state-of-the-art performance on clean data. While backdoor attacks have been thoroughly investigated in the image domain from both attackers' and defenders' sides, an analysis in the frequency domain has been missing thus far. This paper first revisits existing backdoor triggers from a frequency perspective and performs a comprehensive analysis. Our results show that many current backdoor attacks exhibit severe high-frequency artifacts, which persist across different datasets and resolutions. We further demonstrate these high-frequency artifacts enable a simple way to detect existing backdoor triggers at a detection rate of 98.50% without prior knowledge of the attack details and the target model. Acknowledging previous attacks' weaknesses, we propose a practical way to create smooth backdoor triggers without high-frequency artifacts and study their detectability. We show that existing defense works can benefit by incorporating these smooth triggers into their design consideration. Moreover, we show that the detector tuned over stronger smooth triggers can generalize well to unseen weak smooth triggers. In short, our work emphasizes the importance of considering frequency analysis when designing both backdoor attacks and defenses in deep learning.
This paper explores the similarities of output layers in Neural Networks (NNs) with logistic regression to explain importance of inputs by Z-scores. The network analyzed, a network for fusion of Synthetic Aperture Radar (SAR) and Microwave Radiometry (MWR) data, is applied to prediction of arctic sea ice. With the analysis the importance of MWR relative to SAR is found to favor MWR components. Further, as the model represents image features at different scales, the relative importance of these are as well analyzed. The suggested methodology offers a simple and easy framework for analyzing output layer components and can reduce the number of components for further analysis with e.g. common NN visualization methods.
Multimodal automatic speech recognition systems integrate information from images to improve speech recognition quality, by grounding the speech in the visual context. While visual signals have been shown to be useful for recovering entities that have been masked in the audio, these models should be capable of recovering a broader range of word types. Existing systems rely on global visual features that represent the entire image, but localizing the relevant regions of the image will make it possible to recover a larger set of words, such as adjectives and verbs. In this paper, we propose a model that uses finer-grained visual information from different parts of the image, using automatic object proposals. In experiments on the Flickr8K Audio Captions Corpus, we find that our model improves over approaches that use global visual features, that the proposals enable the model to recover entities and other related words, such as adjectives, and that improvements are due to the model's ability to localize the correct proposals.
This paper introduces EfficientNetV2, a new family of convolutional networks that have faster training speed and better parameter efficiency than previous models. To develop this family of models, we use a combination of training-aware neural architecture search and scaling, to jointly optimize training speed and parameter efficiency. The models were searched from the search space enriched with new ops such as Fused-MBConv. Our experiments show that EfficientNetV2 models train much faster than state-of-the-art models while being up to 6.8x smaller. Our training can be further sped up by progressively increasing the image size during training, but it often causes a drop in accuracy. To compensate for this accuracy drop, we propose to adaptively adjust regularization (e.g., dropout and data augmentation) as well, such that we can achieve both fast training and good accuracy. With progressive learning, our EfficientNetV2 significantly outperforms previous models on ImageNet and CIFAR/Cars/Flowers datasets. By pretraining on the same ImageNet21k, our EfficientNetV2 achieves 87.3% top-1 accuracy on ImageNet ILSVRC2012, outperforming the recent ViT by 2.0% accuracy while training 5x-11x faster using the same computing resources. Code will be available at https://github.com/google/automl/efficientnetv2.
Graph convolutional neural network (GCN) has effectively boosted the multi-label image recognition task by introducing label dependencies based on statistical label co-occurrence of data. However, in previous methods, label correlation is computed based on statistical information of data and therefore the same for all samples, and this makes graph inference on labels insufficient to handle huge variations among numerous image instances. In this paper, we propose an instance-aware graph convolutional neural network (IA-GCN) framework for multi-label classification. As a whole, two fused branches of sub-networks are involved in the framework: a global branch modeling the whole image and a region-based branch exploring dependencies among regions of interests (ROIs). For label diffusion of instance-awareness in graph convolution, rather than using the statistical label correlation alone, an image-dependent label correlation matrix (LCM), fusing both the statistical LCM and an individual one of each image instance, is constructed for graph inference on labels to inject adaptive information of label-awareness into the learned features of the model. Specifically, the individual LCM of each image is obtained by mining the label dependencies based on the scores of labels about detected ROIs. In this process, considering the contribution differences of ROIs to multi-label classification, variational inference is introduced to learn adaptive scaling factors for those ROIs by considering their complex distribution. Finally, extensive experiments on MS-COCO and VOC datasets show that our proposed approach outperforms existing state-of-the-art methods.
Nowadays, digital content is widespread and simply redistributable, either lawfully or unlawfully. For example, after images are posted on the internet, other web users can modify them and then repost their versions, thereby generating near-duplicate images. The presence of near-duplicates affects the performance of the search engines critically. Computer vision is concerned with the automatic extraction, analysis and understanding of useful information from digital images. The main application of computer vision is image understanding. There are several tasks in image understanding such as feature extraction, object detection, object recognition, image cleaning, image transformation, etc. There is no proper survey in literature related to near duplicate detection of images. In this paper, we review the state-of-the-art computer vision-based approaches and feature extraction methods for the detection of near duplicate images. We also discuss the main challenges in this field and how other researchers addressed those challenges. This review provides research directions to the fellow researchers who are interested to work in this field.
We tackle a new problem of semantic view synthesis -- generating free-viewpoint rendering of a synthesized scene using a semantic label map as input. We build upon recent advances in semantic image synthesis and view synthesis for handling photographic image content generation and view extrapolation. Direct application of existing image/view synthesis methods, however, results in severe ghosting/blurry artifacts. To address the drawbacks, we propose a two-step approach. First, we focus on synthesizing the color and depth of the visible surface of the 3D scene. We then use the synthesized color and depth to impose explicit constraints on the multiple-plane image (MPI) representation prediction process. Our method produces sharp contents at the original view and geometrically consistent renderings across novel viewpoints. The experiments on numerous indoor and outdoor images show favorable results against several strong baselines and validate the effectiveness of our approach.
Change detection of heterogeneous remote sensing images is an important and challenging topic in remote sensing for emergency situation resulting from nature disaster. Due to the different imaging mechanisms of heterogeneous sensors, it is difficult to directly compare the images. To address this challenge, we explore an unsupervised change detection method based on adaptive local structure consistency (ALSC) between heterogeneous images in this letter, which constructs an adaptive graph representing the local structure for each patch in one image domain and then projects this graph to the other image domain to measure the change level. This local structure consistency exploits the fact that the heterogeneous images share the same structure information for the same ground object, which is imaging modality-invariant. To avoid the leakage of heterogeneous data, the pixelwise change image is calculated in the same image domain by graph projection. Experiment results demonstrate the effectiveness of the proposed ALSC based change detection method by comparing with some state-of-the-art methods.