We propose an image-classification method to predict the perceived-relevance of text documents from eye-movements. An eye-tracking study was conducted where participants read short news articles, and rated them as relevant or irrelevant for answering a trigger question. We encode participants' eye-movement scanpaths as images, and then train a convolutional neural network classifier using these scanpath images. The trained classifier is used to predict participants' perceived-relevance of news articles from the corresponding scanpath images. This method is content-independent, as the classifier does not require knowledge of the screen-content, or the user's information-task. Even with little data, the image classifier can predict perceived-relevance with up to 80% accuracy. When compared to similar eye-tracking studies from the literature, this scanpath image classification method outperforms previously reported metrics by appreciable margins. We also attempt to interpret how the image classifier differentiates between scanpaths on relevant and irrelevant documents.
Text contained in an image carries high-level semantics that can be exploited to achieve richer image understanding. In particular, the mere presence of text provides strong guiding content that should be employed to tackle a diversity of computer vision tasks such as image retrieval, fine-grained classification, and visual question answering. In this paper, we address the problem of fine-grained classification and image retrieval by leveraging textual information along with visual cues to comprehend the existing intrinsic relation between the two modalities. The novelty of the proposed model consists of the usage of a PHOC descriptor to construct a bag of textual words along with a Fisher Vector Encoding that captures the morphology of text. This approach provides a stronger multimodal representation for this task and as our experiments demonstrate, it achieves state-of-the-art results on two different tasks, fine-grained classification and image retrieval.
Foraminifera are single-celled marine organisms, which may have a planktic or benthic lifestyle. During their life cycle they construct shells consisting of one or more chambers, and these shells remain as fossils in marine sediments. Classifying and counting these fossils have become an important tool in e.g. oceanography and climatology. Currently the process of identifying and counting microfossils is performed manually using a microscope and is very time consuming. Developing methods to automate this process is therefore considered important across a range of research fields. The first steps towards developing a deep learning model that can detect and classify microscopic foraminifera are proposed. The proposed model is based on a VGG16 model that has been pretrained on the ImageNet dataset, and adapted to the foraminifera task using transfer learning. Additionally, a novel image dataset consisting of microscopic foraminifera and sediments from the Barents Sea region is introduced.
Compared to supervised learning, semi-supervised learning reduces the dependence of deep learning on a large number of labeled samples. In this work, we use a small number of labeled samples and perform data augmentation on unlabeled samples to achieve image classification. Our method constrains all samples to the predefined evenly-distributed class centroids (PEDCC) by the corresponding loss function. Specifically, the PEDCC-Loss for labeled samples, and the maximum mean discrepancy loss for unlabeled samples are used to make the feature distribution closer to the distribution of PEDCC. Our method ensures that the inter-class distance is large and the intra-class distance is small enough to make the classification boundaries between different classes clearer. Meanwhile, for unlabeled samples, we also use KL divergence to constrain the consistency of the network predictions between unlabeled and augmented samples. Our semi-supervised learning method achieves the state-of-the-art results, with 4000 labeled samples on CIFAR10 and 1000 labeled samples on SVHN, and the accuracy is 95.10% and 97.58% respectively.
Classification of partially occluded images is a highly challenging computer vision problem even for the cutting edge deep learning technologies. To achieve a robust image classification for occluded images, this paper proposes a novel scheme using subspace decomposition based estimation (SDBE). The proposed SDBE-based classification scheme first employs a base convolutional neural network to extract the deep feature vector (DFV) and then utilizes the SDBE to compute the DFV of the original occlusion-free image for classification. The SDBE is performed by projecting the DFV of the occluded image onto the linear span of a class dictionary (CD) along the linear span of an occlusion error dictionary (OED). The CD and OED are constructed respectively by concatenating the DFVs of a training set and the occlusion error vectors of an extra set of image pairs. Two implementations of the SDBE are studied in this paper: the $l_1$-norm and the squared $l_2$-norm regularized least-squares estimates. By employing the ResNet-152, pre-trained on the ILSVRC2012 training set, as the base network, the proposed SBDE-based classification scheme is extensively evaluated on the Caltech-101 and ILSVRC2012 datasets. Extensive experimental results demonstrate that the proposed SDBE-based scheme dramatically boosts the classification accuracy for occluded images, and achieves around $22.25\%$ increase in classification accuracy under $20\%$ occlusion on the ILSVRC2012 dataset.
Retail Product Image Classification is an important Computer Vision and Machine Learning problem for building real world systems like self-checkout stores and automated retail execution evaluation. In this work, we present various tricks to increase accuracy of Deep Learning models on different types of retail product image classification datasets. These tricks enable us to increase the accuracy of fine tuned convnets for retail product image classification by a large margin. As the most prominent trick, we introduce a new neural network layer called Local-Concepts-Accumulation (LCA) layer which gives consistent gains across multiple datasets. Two other tricks we find to increase accuracy on retail product identification are using an instagram-pretrained Convnet and using Maximum Entropy as an auxiliary loss for classification.
A weakly-supervised learning framework named as complementary-label learning has been proposed recently, where each sample is equipped with a single complementary label that denotes one of the classes the sample does not belong to. However, the existing complementary-label learning methods cannot learn from the easily accessible unlabeled samples and samples with multiple complementary labels, which are more informative. In this paper, to remove these limitations, we propose the novel multi-complementary and unlabeled learning framework that allows unbiased estimation of classification risk from samples with any number of complementary labels and unlabeled samples, for arbitrary loss functions and models. We first give an unbiased estimator of the classification risk from samples with multiple complementary labels, and then further improve the estimator by incorporating unlabeled samples into the risk formulation. The estimation error bounds show that the proposed methods are in the optimal parametric convergence rate. Finally, the experiments on both linear and deep models show the effectiveness of our methods.
In recent years, Deep Learning(DL) techniques have been extensively deployed for computer vision tasks, particularly visual classification problems, where new algorithms reported to achieve or even surpass the human performance. While many recent works demonstrated that DL models are vulnerable to adversarial examples. Fortunately, generating adversarial examples usually requires white-box access to the victim model, and real-world cloud-based image classification services are more complex than white-box classifier,the architecture and parameters of DL models on cloud platforms cannot be obtained by the attacker. The attacker can only access the APIs opened by cloud platforms. Thus, keeping models in the cloud can usually give a (false) sense of security. In this paper, we mainly focus on studying the security of real-world cloud-based image classification services. Specifically, (1) We propose a novel attack method, Fast Featuremap Loss PGD (FFL-PGD) attack based on Substitution model, which achieves a high bypass rate with a very limited number of queries. Instead of millions of queries in previous studies, our method finds the adversarial examples using only two queries per image; and (2) we make the first attempt to conduct an extensive empirical study of black-box attacks against real-world cloud-based classification services. Through evaluations on four popular cloud platforms including Amazon, Google, Microsoft, Clarifai, we demonstrate that FFL-PGD attack has a success rate over 90\% among different classification services. (3) We discuss the possible defenses to address these security challenges in cloud-based classification services. Our defense technology is mainly divided into model training stage and image preprocessing stage.
Colorectal cancer, largely arising from precursor lesions called polyps, remains one of the leading causes of cancer-related death worldwide. Current clinical standards require the resection and histopathological analysis of polyps due to test accuracy and sensitivity of optical biopsy methods falling substantially below recommended levels. In this study, we design a novel capsule network architecture (D-Caps) to improve the viability of optical biopsy of colorectal polyps. Our proposed method introduces several technical novelties including a novel capsule architecture with a capsule-average pooling (CAP) method to improve efficiency in large-scale image classification. We demonstrate improved results over the previous state-of-the-art convolutional neural network (CNN) approach by as much as 43%. This work provides an important benchmark on the new Mayo Polyp dataset, a significantly more challenging and larger dataset than previous polyp studies, with results stratified across all available categories, imaging devices and modalities, and focus modes to promote future direction into AI-driven colorectal cancer screening systems. Code is publicly available at https://github.com/lalonderodney/D-Caps .
Transfer learning has proven to be a successful technique to train deep learning models in the domains where little training data is available. The dominant approach is to pretrain a model on a large generic dataset such as ImageNet and finetune its weights on the target domain. However, in the new era of an ever-increasing number of massive datasets, selecting the relevant data for pretraining is a critical issue. We introduce Neural Data Server (NDS), a large-scale search engine for finding the most useful transfer learning data to the target domain. NDS consists of a dataserver which indexes several large popular image datasets, and aims to recommend data to a client, an end-user with a target application with its own small labeled dataset. The dataserver represents large datasets with a much more compact mixture-of-experts model, and employs it to perform data search in a series of dataserver-client transactions at a low computational cost. We show the effectiveness of NDS in various transfer learning scenarios, demonstrating state-of-the-art performance on several target datasets and tasks such as image classification, object detection and instance segmentation. Neural Data Server is available as a web-service at http://aidemos.cs.toronto.edu/nds/.
Deep neural networks achieve remarkable performance in many computer vision tasks. Most state-of-the-art (SOTA) semantic segmentation and object detection approaches reuse neural network architectures designed for image classification as the backbone, commonly pre-trained on ImageNet. However, performance gains can be achieved by designing network architectures specifically for detection and segmentation, as shown by recent neural architecture search (NAS) research for detection and segmentation. One major challenge though, is that ImageNet pre-training of the search space representation (a.k.a. super network) or the searched networks incurs huge computational cost. In this paper, we propose a Fast Neural Network Adaptation (FNA) method, which can adapt both the architecture and parameters of a seed network (e.g. a high performing manually designed backbone) to become a network with different depth, width, or kernels via a Parameter Remapping technique, making it possible to utilize NAS for detection/segmentation tasks a lot more efficiently. In our experiments, we conduct FNA on MobileNetV2 to obtain new networks for both segmentation and detection that clearly out-perform existing networks designed both manually and by NAS. The total computation cost of FNA is significantly less than SOTA segmentation/detection NAS approaches: 1737$\times$ less than DPC, 6.8$\times$ less than Auto-DeepLab and 7.4$\times$ less than DetNAS. The code is available at https://github.com/JaminFong/FNA.
In the last years, image classification processes like neural networks in the area of art-history and Heritage Informatics have experienced a broad distribution (Lang and Ommer 2018). These methods face several challenges, including the handling of comparatively small amounts of data as well as high-dimensional data in the Digital Humanities. Here, a Convolutional Neural Network (CNN) is used that output is not, as usual, a series of flat text labels but a series of semantically loaded vectors. These vectors result from a Distributional Semantic Model (DSM) which is generated from an in-domain text corpus. ----- In den letzten Jahren hat die Verwendung von Bildklassifizierungsverfahren wie neuronalen Netzwerken auch im Bereich der historischen Bildwissenschaften und der Heritage Informatics weite Verbreitung gefunden (Lang und Ommer 2018). Diese Verfahren stehen dabei vor einer Reihe von Herausforderungen, darunter dem Umgangmit den vergleichsweise kleinen Datenmengen sowie zugleich hochdimensionalen Da-tenr\"aumen in den digitalen Geisteswissenschaften. Meist bilden diese Methoden dieKlassifizierung auf einen vergleichsweise flachen Raum ab. Dieser flache Zugang verliert im Bem\"uhen um ontologische Eindeutigkeit eine Reihe von relevanten Dimensionen, darunter taxonomische, mereologische und assoziative Beziehungen zwischenden Klassen beziehungsweise dem nicht formalisierten Kontext. Dabei wird ein Convolutional Neural Network (CNN) genutzt, dessen Ausgabe im Trainingsprozess, anders als herk\"ommlich, nicht auf einer Serie flacher Textlabel beruht, sondern auf einer Serie von Vektoren. Diese Vektoren resultieren aus einem Distributional Semantic Model (DSM), welches aus einem Dom\"ane-Textkorpus generiert wird.
Fine-grained image classification is a challenging problem, since the difficulty of finding discriminative features. To handle this circumstance, basically, there are two ways to go. One is use attention based method to focus on informative areas, while the other one aims to find high order between features. Further, for attention based method there are two directions, activation based and detection based, which are proved effective by scholars. However ,rare work focus on fusing two types of attention with high order feature. In this paper, we propose a novel DAF method which fuse two types of attention and use them to as PAF(part attention filter) in deep bilinear transformation module to mine the relationship between separate parts of an object. Briefly, our network constructed by a student net who attempt to output two attention maps and a teacher net uses these two maps as empirical information to refine the result. The experiment result shows that only student net could get 87.6% accuracy in CUB dataset while cooperating with teacher net could achieve 89.1% accuracy.
Neural network training is computationally and memory intensive. Sparse training can reduce the burden on emerging hardware platforms designed to accelerate sparse computations, but it can affect network convergence. In this work, we propose a novel CNN training algorithm Sparse Weight Activation Training (SWAT). SWAT is more computation and memory-efficient than conventional training. SWAT modifies back-propagation based on the empirical insight that convergence during training tends to be robust to the elimination of (i) small magnitude weights during the forward pass and (ii) both small magnitude weights and activations during the backward pass. We evaluate SWAT on recent CNN architectures such as ResNet, VGG, DenseNet and WideResNet using CIFAR-10, CIFAR-100 and ImageNet datasets. For ResNet-50 on ImageNet SWAT reduces total floating-point operations (FLOPS) during training by 80% resulting in a 3.3$\times$ training speedup when run on a simulated sparse learning accelerator representative of emerging platforms while incurring only 1.63% reduction in validation accuracy. Moreover, SWAT reduces memory footprint during the backward pass by 23% to 50% for activations and 50% to 90% for weights.
The combination of neural network and fuzzy systems into neuro-fuzzy systems integrates fuzzy reasoning rules into the connectionist networks. However, the existing neuro-fuzzy systems are developed under shallow structures having lower generalization capacity. We propose the first end-to-end deep neuro-fuzzy network and investigate its application for image classification. Two new operations are developed based on definitions of Takagi-Sugeno-Kang (TSK) fuzzy model namely fuzzy inference operation and fuzzy pooling operations; stacks of these operations comprise the layers in this network. We evaluate the network on MNIST, CIFAR-10 and CIFAR-100 datasets, finding that the network has a reasonable accuracy in these benchmarks.
In order to port the performance of trained artificial neural networks (ANNs) to spiking neural networks (SNNs), which can be implemented in neuromorphic hardware with a drastically reduced energy consumption, an efficient ANN to SNN conversion is needed. Previous conversion schemes focused on the representation of the analog output of a rectified linear (ReLU) gate in the ANN by the firing rate of a spiking neuron. But this is not possible for other commonly used ANN gates, and it reduces the throughput even for ReLU gates. We introduce a new conversion method where a gate in the ANN, which can basically be of any type, is emulated by a small circuit of spiking neurons, with At Most One Spike (AMOS) per neuron. We show that this AMOS conversion improves the accuracy of SNNs for ImageNet from 74.60% to 80.97%, thereby bringing it within reach of the best available ANN accuracy (85.0%). The Top5 accuracy of SNNs is raised to 95.82%, getting even closer to the best Top5 performance of 97.2% for ANNs. In addition, AMOS conversion improves latency and throughput of spike-based image classification by several orders of magnitude. Hence these results suggest that SNNs provide a viable direction for developing highly energy efficient hardware for AI that combines high performance with versatility of applications.
Spiking Neural Networks (SNNs) are brain-inspired, event-driven machine learning algorithms that have been widely recognized in producing ultra-high-energy-efficient hardware. Among existing SNNs, unsupervised SNNs based on synaptic plasticity, especially Spike-Timing-Dependent Plasticity (STDP), are considered to have great potential in imitating the learning process of the biological brain. Nevertheless, the existing STDP-based SNNs have limitations in constrained learning capability and/or slow learning speed. Most STDP-based SNNs adopted a slow-learning Fully-Connected (FC) architecture and used a sub-optimal vote-based scheme for spike decoding. In this paper, we overcome these limitations with: 1) a design of high-parallelism network architecture, inspired by the Inception module in Artificial Neural Networks (ANNs); 2) use of a Vote-for-All (VFA) decoding layer as a replacement to the standard vote-based spike decoding scheme, to reduce the information loss in spike decoding and, 3) a proposed adaptive repolarization (resetting) mechanism that accelerates SNNs' learning by enhancing spiking activities. Our experimental results on two established benchmark datasets (MNIST/EMNIST) show that our network architecture resulted in superior performance compared to the widely used FC architecture and a more advanced Locally-Connected (LC) architecture, and that our SNN achieved competitive results with state-of-the-art unsupervised SNNs (95.64%/80.11% accuracy on the MNIST/EMNISE dataset) while having superior learning efficiency and robustness against hardware damage. Our SNN achieved great classification accuracy with only hundreds of training iterations, and random destruction of large numbers of synapses or neurons only led to negligible performance degradation.
Deep neural networks (DNNs) have achieved impressive performance on handling computer vision problems, however, it has been found that DNNs are vulnerable to adversarial examples. For such reason, adversarial perturbations have been recently studied in several respects. However, most previous works have focused on image classification tasks, and it has never been studied regarding adversarial perturbations on Image-to-image (Im2Im) translation tasks, showing great success in handling paired and/or unpaired mapping problems in the field of autonomous driving and robotics. This paper examines different types of adversarial perturbations that can fool Im2Im frameworks for autonomous driving purpose. We propose both quasi-physical and digital adversarial perturbations that can make Im2Im models yield unexpected results. We then empirically analyze these perturbations and show that they generalize well under both paired for image synthesis and unpaired settings for style transfer. We also validate that there exist some perturbation thresholds over which the Im2Im mapping is disrupted or impossible. The existence of these perturbations reveals that there exist crucial weaknesses in Im2Im models. Lastly, we show that our methods illustrate how perturbations affect the quality of outputs, pioneering the improvement of the robustness of current SOTA networks for autonomous driving.
Creating fake images and videos such as "Deepfake" has become much easier these days due to the advancement in Generative Adversarial Networks (GANs). Moreover, recent research such as the few-shot learning can create highly realistic personalized fake images with only a few images. Therefore, the threat of Deepfake to be used for a variety of malicious intents such as propagating fake images and videos becomes prevalent. And detecting these machine-generated fake images has been quite challenging than ever. In this work, we propose a light-weight robust fine-tuning neural network-based classifier architecture called Fake Detection Fine-tuning Network (FDFtNet), which is capable of detecting many of the new fake face image generation models, and can be easily combined with existing image classification networks and finetuned on a few datasets. In contrast to many existing methods, our approach aims to reuse popular pre-trained models with only a few images for fine-tuning to effectively detect fake images. The core of our approach is to introduce an image-based self-attention module called Fine-Tune Transformer that uses only the attention module and the down-sampling layer. This module is added to the pre-trained model and fine-tuned on a few data to search for new sets of feature space to detect fake images. We experiment with our FDFtNet on the GANsbased dataset (Progressive Growing GAN) and Deepfake-based dataset (Deepfake and Face2Face) with a small input image resolution of 64x64 that complicates detection. Our FDFtNet achieves an overall accuracy of 90.29% in detecting fake images generated from the GANs-based dataset, outperforming the state-of-the-art.
Methods inspired by Artificial Intelligence (AI) are starting to fundamentally change computational science and engineering through breakthrough performances on challenging problems. However, reliability and trustworthiness of such techniques is a major concern. In inverse problems in imaging, the focus of this paper, there is increasing empirical evidence that methods may suffer from hallucinations, i.e., false, but realistic-looking artifacts; instability, i.e., sensitivity to perturbations in the data; and unpredictable generalization, i.e., excellent performance on some images, but significant deterioration on others. This paper provides a theoretical foundation for these phenomena. We give mathematical explanations for how and when such effects arise in arbitrary reconstruction methods, with several of our results taking the form of `no free lunch' theorems. Specifically, we show that (i) methods that overperform on a single image can wrongly transfer details from one image to another, creating a hallucination, (ii) methods that overperform on two or more images can hallucinate or be unstable, (iii) optimizing the accuracy-stability trade-off is generally difficult, (iv) hallucinations and instabilities, if they occur, are not rare events, and may be encouraged by standard training, (v) it may be impossible to construct optimal reconstruction maps for certain problems. Our results trace these effects to the kernel of the forward operator whenever it is nontrivial, but also apply to the case when the forward operator is ill-conditioned. Based on these insights, our work aims to spur research into new ways to develop robust and reliable AI-based methods for inverse problems in imaging.