1.Self-supervised Modal and View Invariant Feature Learning ⬇️
Most of the existing self-supervised feature learning methods for 3D data either learn 3D features from point cloud data or from multi-view images. By exploring the inherent multi-modality attributes of 3D objects, in this paper, we propose to jointly learn modal-invariant and view-invariant features from different modalities including image, point cloud, and mesh with heterogeneous networks for 3D data. In order to learn modal- and view-invariant features, we propose two types of constraints: cross-modal invariance constraint and cross-view invariant constraint. Cross-modal invariance constraint forces the network to maximum the agreement of features from different modalities for same objects, while the cross-view invariance constraint forces the network to maximum agreement of features from different views of images for same objects. The quality of learned features has been tested on different downstream tasks with three modalities of data including point cloud, multi-view images, and mesh. Furthermore, the invariance cross different modalities and views are evaluated with the cross-modal retrieval task. Extensive evaluation results demonstrate that the learned features are robust and have strong generalizability across different tasks.
2.Modeling the Distribution of Normal Data in Pre-Trained Deep Features for Anomaly Detection ⬇️
Anomaly Detection (AD) in images is a fundamental computer vision problem and refers to identifying images and/or image substructures that deviate significantly from the norm. Popular AD algorithms commonly try to learn a model of normality from scratch using task specific datasets, but are limited to semi-supervised approaches employing mostly normal data due to the inaccessibility of anomalies on a large scale combined with the ambiguous nature of anomaly appearance.
We follow an alternative approach and demonstrate that deep feature representations learned by discriminative models on large natural image datasets are well suited to describe normality and detect even subtle anomalies. Our model of normality is established by fitting a multivariate Gaussian to deep feature representations of classification networks trained on ImageNet using normal data only in a transfer learning setting. By subsequently applying the Mahalanobis distance as the anomaly score we outperform the current state of the art on the public MVTec AD dataset, achieving an Area Under the Receiver Operating Characteristic curve of$95.8 \pm 1.2$ (mean$\pm$ SEM) over all 15 classes. We further investigate why the learned representations are discriminative to the AD task using Principal Component Analysis. We find that the principal components containing little variance in normal data are the ones crucial for discriminating between normal and anomalous instances. This gives a possible explanation to the often sub-par performance of AD approaches trained from scratch using normal data only. By selectively fitting a multivariate Gaussian to these most relevant components only we are able to further reduce model complexity while retaining AD performance. We also investigate setting the working point by selecting acceptable False Positive Rate thresholds based on the multivariate Gaussian assumption.
3.Unsupervised learning of multimodal image registration using domain adaptation with projected Earth Move's discrepancies ⬇️
Multimodal image registration is a very challenging problem for deep learning approaches. Most current work focuses on either supervised learning that requires labelled training scans and may yield models that bias towards annotated structures or unsupervised approaches that are based on hand-crafted similarity metrics and may therefore not outperform their classical non-trained counterparts. We believe that unsupervised domain adaptation can be beneficial in overcoming the current limitations for multimodal registration, where good metrics are hard to define. Domain adaptation has so far been mainly limited to classification problems. We propose the first use of unsupervised domain adaptation for discrete multimodal registration. Based on a source domain for which quantised displacement labels are available as supervision, we transfer the output distribution of the network to better resemble the target domain (other modality) using classifier discrepancies. To improve upon the sliced Wasserstein metric for 2D histograms, we present a novel approximation that projects predictions into 1D and computes the L1 distance of their cumulative sums. Our proof-of-concept demonstrates the applicability of domain transfer from mono- to multimodal (multi-contrast) 2D registration of canine MRI scans and improves the registration accuracy from 33% (using sliced Wasserstein) to 44%.
4.Uncertainty-Aware Blind Image Quality Assessment in the Laboratory and Wild ⬇️
Performance of blind image quality assessment (BIQA) models has been significantly boosted by end-to-end optimization of feature engineering and quality regression. Nevertheless, due to the distributional shifts between images simulated in the laboratory and captured in the wild, models trained on databases with synthetic distortions remain particularly weak at handling realistic distortions (and vice versa). To confront the cross-distortion-scenario challenge, we develop a unified BIQA model and an effective approach of training it for both synthetic and realistic distortions. We first sample pairs of images from the same IQA databases and compute a probability that one image of each pair is of higher quality as the supervisory signal. We then employ the fidelity loss to optimize a deep neural network for BIQA over a large number of such image pairs. We also explicitly enforce a hinge constraint to regularize uncertainty estimation during optimization. Extensive experiments on six IQA databases show the promise of the learned method in blindly assessing image quality in the laboratory and wild. In addition, we demonstrate the universality of the proposed training strategy by using it to improve existing BIQA models.
5.Robust Modeling of Epistemic Mental States ⬇️
This work identifies and advances some research challenges in the analysis of facial features and their temporal dynamics with epistemic mental states in dyadic conversations. Epistemic states are: Agreement, Concentration, Thoughtful, Certain, and Interest. In this paper, we perform a number of statistical analyses and simulations to identify the relationship between facial features and epistemic states. Non-linear relations are found to be more prevalent, while temporal features derived from original facial features have demonstrated a strong correlation with intensity changes. Then, we propose a novel prediction framework that takes facial features and their nonlinear relation scores as input and predict different epistemic states in videos. The prediction of epistemic states is boosted when the classification of emotion changing regions such as rising, falling, or steady-state are incorporated with the temporal features. The proposed predictive models can predict the epistemic states with significantly improved accuracy: correlation coefficient (CoERR) for Agreement is 0.827, for Concentration 0.901, for Thoughtful 0.794, for Certain 0.854, and for Interest 0.913.
6.Improving Generalized Zero-Shot Learning by Semantic Discriminator ⬇️
It is a recognized fact that the classification accuracy of unseen classes in the setting of Generalized Zero-Shot Learning (GZSL) is much lower than that of traditional Zero-Shot Leaning (ZSL). One of the reasons is that an instance is always misclassified to the wrong domain. Here we refer to the seen and unseen classes as two domains respectively. We propose a new approach to distinguish whether the instances come from the seen or unseen classes. First the visual feature of instance is projected into the semantic space. Then the absolute norm difference between the projected semantic vector and the class semantic embedding vector, and the minimum distance between the projected semantic vectors and the semantic embedding vectors of the seen classes are used as discrimination basis. This approach is termed as SD (Semantic Discriminator) because domain judgement of instance is performed in the semantic space. Our approach can be combined with any existing ZSL method and fully supervision classification model to form a new GZSL method. Furthermore, our approach is very simple and does not need any fixed parameters. A large number of experiments show that the accuracy of our approach is 8.5% to 21.9% higher than the current best method.
7.Disentanglement Then Reconstruction: Learning Compact Features for Unsupervised Domain Adaptation ⬇️
Recent works in domain adaptation always learn domain invariant features to mitigate the gap between the source and target domains by adversarial methods. The category information are not sufficiently used which causes the learned domain invariant features are not enough discriminative. We propose a new domain adaptation method based on prototype construction which likes capturing data cluster centers. Specifically, it consists of two parts: disentanglement and reconstruction. First, the domain specific features and domain invariant features are disentangled from the original features. At the same time, the domain prototypes and class prototypes of both domains are estimated. Then, a reconstructor is trained by reconstructing the original features from the disentangled domain invariant features and domain specific features. By this reconstructor, we can construct prototypes for the original features using class prototypes and domain prototypes correspondingly. In the end, the feature extraction network is forced to extract features close to these prototypes. Our contribution lies in the technical use of the reconstructor to obtain the original feature prototypes which helps to learn compact and discriminant features. As far as we know, this idea is proposed for the first time. Experiment results on several public datasets confirm the state-of-the-art performance of our method.
8.Quantifying the Complexity of Standard Benchmarking Datasets for Long-Term Human Trajectory Prediction ⬇️
Methods to quantify the complexity of trajectory datasets are still a missing piece in benchmarking human trajectory prediction models. In order to gain a better understanding of the complexity of trajectory datasets, an approach for deriving complexity scores from a prototype-based dataset representation is proposed. The dataset representation is obtained by first employing a non-trivial spatial sequence alignment, which enables a following learning vector quantization (LVQ) stage. A large-scale complexity analysis is conducted on several human trajectory prediction benchmarking datasets, followed by a brief discussion on indications for human trajectory prediction and benchmarking.
9.CNN-based Approach for Cervical Cancer Classification in Whole-Slide Histopathology Images ⬇️
Cervical cancer will cause 460 000 deaths per year by 2040, approximately 90% are Sub-Saharan African women. A constantly increasing incidence in Africa making cervical cancer a priority by the World Health Organization (WHO) in terms of screening, diagnosis, and treatment. Conventionally, cancer diagnosis relies primarily on histopathological assessment, a deeply error-prone procedure requiring intelligent computer-aided systems as low-cost patient safety mechanisms but lack of labeled data in digital pathology limits their applicability. In this study, few cervical tissue digital slides from TCGA data portal were pre-processed to overcome whole-slide images obstacles and included in our proposed VGG16-CNN classification approach. Our results achieved an accuracy of 98,26% and an F1-score of 97,9%, which confirm the potential of transfer learning on this weakly-supervised task.
10.P2B: Point-to-Box Network for 3D Object Tracking in Point Clouds ⬇️
Towards 3D object tracking in point clouds, a novel point-to-box network termed P2B is proposed in an end-to-end learning manner. Our main idea is to first localize potential target centers in 3D search area embedded with target information. Then point-driven 3D target proposal and verification are executed jointly. In this way, the time-consuming 3D exhaustive search can be avoided. Specifically, we first sample seeds from the point clouds in template and search area respectively. Then, we execute permutation-invariant feature augmentation to embed target clues from template into search area seeds and represent them with target-specific features. Consequently, the augmented search area seeds regress the potential target centers via Hough voting. The centers are further strengthened with seed-wise targetness scores. Finally, each center clusters its neighbors to leverage the ensemble power for joint 3D target proposal and verification. We apply PointNet++ as our backbone and experiments on KITTI tracking dataset demonstrate P2B's superiority (~10%'s improvement over state-of-the-art). Note that P2B can run with 40FPS on a single NVIDIA 1080Ti GPU. Our code and model are available at this https URL.
11.CGGAN: A Context Guided Generative Adversarial Network For Single Image Dehazing ⬇️
Image haze removal is highly desired for the application of computer vision. This paper proposes a novel Context Guided Generative Adversarial Network (CGGAN) for single image dehazing. Of which, an novel new encoder-decoder is employed as the generator. And it consists of a feature-extraction-net, a context-extractionnet, and a fusion-net in sequence. The feature extraction-net acts as a encoder, and is used for extracting haze features. The context-extraction net is a multi-scale parallel pyramid decoder, and is used for extracting the deep features of the encoder and generating coarse dehazing image. The fusion-net is a decoder, and is used for obtaining the final haze-free image. To obtain more better results, multi-scale information obtained during the decoding process of the context extraction decoder is used for guiding the fusion decoder. By introducing an extra coarse decoder to the original encoder-decoder, the CGGAN can make better use of the deep feature information extracted by the encoder. To ensure our CGGAN work effectively for different haze scenarios, different loss functions are employed for the two decoders. Experiments results show the advantage and the effectiveness of our proposed CGGAN, evidential improvements over existing state-of-the-art methods are obtained.
12.Traditional Method Inspired Deep Neural Network for Edge Detection ⬇️
Recently, Deep-Neural-Network (DNN) based edge prediction is progressing fast. Although the DNN based schemes outperform the traditional edge detectors, they have much higher computational complexity. It could be that the DNN based edge detectors often adopt the neural net structures designed for high-level computer vision tasks, such as image segmentation and object recognition. Edge detection is a rather local and simple job, the over-complicated architecture and massive parameters may be unnecessary. Therefore, we propose a traditional method inspired framework to produce good edges with minimal complexity. We simplify the network architecture to include Feature Extractor, Enrichment, and Summarizer, which roughly correspond to gradient, low pass filter, and pixel connection in the traditional edge detection schemes. The proposed structure can effectively reduce the complexity and retain the edge prediction quality. Our TIN2 (Traditional Inspired Network) model has an accuracy higher than the recent BDCN2 (Bi-Directional Cascade Network) but with a smaller model.
13.Boosting Few-Shot Learning With Adaptive Margin Loss ⬇️
Few-shot learning (FSL) has attracted increasing attention in recent years but remains challenging, due to the intrinsic difficulty in learning to generalize from a few examples. This paper proposes an adaptive margin principle to improve the generalization ability of metric-based meta-learning approaches for few-shot learning problems. Specifically, we first develop a class-relevant additive margin loss, where semantic similarity between each pair of classes is considered to separate samples in the feature embedding space from similar classes. Further, we incorporate the semantic context among all classes in a sampled training task and develop a task-relevant additive margin loss to better distinguish samples from different classes. Our adaptive margin method can be easily extended to a more realistic generalized FSL setting. Extensive experiments demonstrate that the proposed method can boost the performance of current metric-based meta-learning approaches, under both the standard FSL and generalized FSL settings.
14.TOAN: Target-Oriented Alignment Network for Fine-Grained Image Categorization with Few Labeled Samples ⬇️
The challenges of high intra-class variance yet low inter-class fluctuations in fine-grained visual categorization are more severe with few labeled samples, \textit{i.e.,} Fine-Grained categorization problems under the Few-Shot setting (FGFS). High-order features are usually developed to uncover subtle differences between sub-categories in FGFS, but they are less effective in handling the high intra-class variance. In this paper, we propose a Target-Oriented Alignment Network (TOAN) to investigate the fine-grained relation between the target query image and support classes. The feature of each support image is transformed to match the query ones in the embedding feature space, which reduces the disparity explicitly within each category. Moreover, different from existing FGFS approaches devise the high-order features over the global image with less explicit consideration of discriminative parts, we generate discriminative fine-grained features by integrating compositional concept representations to global second-order pooling. Extensive experiments are conducted on four fine-grained benchmarks to demonstrate the effectiveness of TOAN compared with the state-of-the-art models.
15.Explainable deep learning models in medical image analysis ⬇️
Deep learning methods have been very effective for a variety of medical diagnostic tasks and has even beaten human experts on some of those. However, the black-box nature of the algorithms has restricted clinical use. Recent explainability studies aim to show the features that influence the decision of a model the most. The majority of literature reviews of this area have focused on taxonomy, ethics, and the need for explanations. A review of the current applications of explainable deep learning for different medical imaging tasks is presented here. The various approaches, challenges for clinical deployment, and the areas requiring further research are discussed here from a practical standpoint of a deep learning researcher designing a system for the clinical end-users.
16.3D human pose estimation with adaptive receptive fields and dilated temporal convolutions ⬇️
In this work, we demonstrate that receptive fields in 3D pose estimation can be effectively specified using optical flow. We introduce adaptive receptive fields, a simple and effective method to aid receptive field selection in pose estimation models based on optical flow inference. We contrast the performance of a benchmark state-of-the-art model running on fixed receptive fields with their adaptive field counterparts. By using a reduced receptive field, our model can process slow-motion sequences (10x longer) 23% faster than the benchmark model running at regular speed. The reduction in computational cost is achieved while producing a pose prediction accuracy to within 0.36% of the benchmark model.
17.Stereo Vision Based Single-Shot 6D Object Pose Estimation for Bin-Picking by a Robot Manipulator ⬇️
We propose a fast and accurate method of 6D object pose estimation for bin-picking of mechanical parts by a robot manipulator. We extend the single-shot approach to stereo vision by application of attention architecture. Our convolutional neural network model regresses to object locations and rotations from either a left image or a right image without depth information. Then, a stereo feature matching module, designated as Stereo Grid Attention, generates stereo grid matching maps. The important point of our method is only to calculate disparity of the objects found by the attention from stereo images, instead of calculating a point cloud over the entire image. The disparity value is then used to calculate the depth to the objects by the principle of triangulation. Our method also achieves a rapid processing speed of pose estimation by the single-shot architecture and it is possible to process a 1024 x 1024 pixels image in 75 milliseconds on the Jetson AGX Xavier implemented with half-float model. Weakly textured mechanical parts are used to exemplify the method. First, we create original synthetic datasets for training and evaluating of the proposed model. This dataset is created by capturing and rendering numerous 3D models of several types of mechanical parts in virtual space. Finally, we use a robotic manipulator with an electromagnetic gripper to pick up the mechanical parts in a cluttered state to verify the validity of our method in an actual scene. When a raw stereo image is used by the proposed method from our stereo camera to detect black steel screws, stainless screws, and DC motor parts, i.e., cases, rotor cores and commutator caps, the bin-picking tasks are successful with 76.3%, 64.0%, 50.5%, 89.1% and 64.2% probability, respectively.
18.Universal Lesion Detection by Learning from Multiple Heterogeneously Labeled Datasets ⬇️
Lesion detection is an important problem within medical imaging analysis. Most previous work focuses on detecting and segmenting a specialized category of lesions (e.g., lung nodules). However, in clinical practice, radiologists are responsible for finding all possible types of anomalies. The task of universal lesion detection (ULD) was proposed to address this challenge by detecting a large variety of lesions from the whole body. There are multiple heterogeneously labeled datasets with varying label completeness: DeepLesion, the largest dataset of 32,735 annotated lesions of various types, but with even more missing annotation instances; and several fully-labeled single-type lesion datasets, such as LUNA for lung nodules and LiTS for liver tumors. In this work, we propose a novel framework to leverage all these datasets together to improve the performance of ULD. First, we learn a multi-head multi-task lesion detector using all datasets and generate lesion proposals on DeepLesion. Second, missing annotations in DeepLesion are retrieved by a new method of embedding matching that exploits clinical prior knowledge. Last, we discover suspicious but unannotated lesions using knowledge transfer from single-type lesion detectors. In this way, reliable positive and negative regions are obtained from partially-labeled and unlabeled images, which are effectively utilized to train ULD. To assess the clinically realistic protocol of 3D volumetric ULD, we fully annotated 1071 CT sub-volumes in DeepLesion. Our method outperforms the current state-of-the-art approach by 29% in the metric of average sensitivity.
19.L^2UWE: A Framework for the Efficient Enhancement of Low-Light Underwater Images Using Local Contrast and Multi-Scale Fusion ⬇️
Images captured underwater often suffer from suboptimal illumination settings that can hide important visual features, reducing their quality. We present a novel single-image low-light underwater image enhancer, L^2UWE, that builds on our observation that an efficient model of atmospheric lighting can be derived from local contrast information. We create two distinct models and generate two enhanced images from them: one that highlights finer details, the other focused on darkness removal. A multi-scale fusion process is employed to combine these images while emphasizing regions of higher luminance, saliency and local contrast. We demonstrate the performance of L^2UWE by using seven metrics to test it against seven state-of-the-art enhancement methods specific to underwater and low-light scenes.
20.Anomaly Detection Based on Deep Learning Using Video for Prevention of Industrial Accidents ⬇️
This paper proposes an anomaly detection method for the prevention of industrial accidents using machine learning technology.
21.Few-Shot Open-Set Recognition using Meta-Learning ⬇️
The problem of open-set recognition is considered. While previous approaches only consider this problem in the context of large-scale classifier training, we seek a unified solution for this and the low-shot classification setting. It is argued that the classic softmax classifier is a poor solution for open-set recognition, since it tends to overfit on the training classes. Randomization is then proposed as a solution to this problem. This suggests the use of meta-learning techniques, commonly used for few-shot classification, for the solution of open-set recognition. A new oPen sEt mEta LEaRning (PEELER) algorithm is then introduced. This combines the random selection of a set of novel classes per episode, a loss that maximizes the posterior entropy for examples of those classes, and a new metric learning formulation based on the Mahalanobis distance. Experimental results show that PEELER achieves state of the art open set recognition performance for both few-shot and large-scale recognition. On CIFAR and miniImageNet, it achieves substantial gains in seen/unseen class detection AUROC for a given seen-class classification accuracy.
22.AFAT: Adaptive Failure-Aware Tracker for Robust Visual Object Tracking ⬇️
Siamese approaches have achieved promising performance in visual object tracking recently. The key to the success of Siamese trackers is to learn appearance-invariant feature embedding functions via pair-wise offline training on large-scale video datasets. However, the Siamese paradigm uses one-shot learning to model the online tracking task, which impedes online adaptation in the tracking process. Additionally, the uncertainty of an online tracking response is not measured, leading to the problem of ignoring potential failures. In this paper, we advocate online adaptation in the tracking stage. To this end, we propose a failure-aware system, realised by a Quality Prediction Network (QPN), based on convolutional and LSTM modules in the decision stage, enabling online reporting of potential tracking failures. Specifically, sequential response maps from previous successive frames as well as current frame are collected to predict the tracking confidence, realising spatio-temporal fusion in the decision level. In addition, we further provide an Adaptive Failure-Aware Tracker (AFAT) by combing the state-of-the-art Siamese trackers with our system. The experimental results obtained on standard benchmarking datasets demonstrate the effectiveness of the proposed failure-aware system and the merits of our AFAT tracker, with outstanding and balanced performance in both accuracy and speed.
23.Detecting Scatteredly-Distributed, Small, andCritically Important Objects in 3D OncologyImaging via Decision Stratification ⬇️
Finding and identifying scatteredly-distributed, small, and critically important objects in 3D oncology images is very challenging. We focus on the detection and segmentation of oncology-significant (or suspicious cancer metastasized) lymph nodes (OSLNs), which has not been studied before as a computational task. Determining and delineating the spread of OSLNs is essential in defining the corresponding resection/irradiating regions for the downstream workflows of surgical resection and radiotherapy of various cancers. For patients who are treated with radiotherapy, this task is performed by experienced radiation oncologists that involves high-level reasoning on whether LNs are metastasized, which is subject to high inter-observer variations. In this work, we propose a divide-and-conquer decision stratification approach that divides OSLNs into tumor-proximal and tumor-distal categories. This is motivated by the observation that each category has its own different underlying distributions in appearance, size and other characteristics. Two separate detection-by-segmentation networks are trained per category and fused. To further reduce false positives (FP), we present a novel global-local network (GLNet) that combines high-level lesion characteristics with features learned from localized 3D image patches. Our method is evaluated on a dataset of 141 esophageal cancer patients with PET and CT modalities (the largest to-date). Our results significantly improve the recall from
$45%$ to$67%$ at$3$ FPs per patient as compared to previous state-of-the-art methods. The highest achieved OSLN recall of$0.828$ is clinically relevant and valuable.
24.D2D: Keypoint Extraction with Describe to Detect Approach ⬇️
In this paper, we present a novel approach that exploits the information within the descriptor space to propose keypoint locations. Detect then describe, or detect and describe jointly are two typical strategies for extracting local descriptors. In contrast, we propose an approach that inverts this process by first describing and then detecting the keypoint locations. % Describe-to-Detect (D2D) leverages successful descriptor models without the need for any additional training. Our method selects keypoints as salient locations with high information content which is defined by the descriptors rather than some independent operators. We perform experiments on multiple benchmarks including image matching, camera localisation, and 3D reconstruction. The results indicate that our method improves the matching performance of various descriptors and that it generalises across methods and tasks.
25.Network Fusion for Content Creation with Conditional INNs ⬇️
Artificial Intelligence for Content Creation has the potential to reduce the amount of manual content creation work significantly. While automation of laborious work is welcome, it is only useful if it allows users to control aspects of the creative process when desired. Furthermore, widespread adoption of semi-automatic content creation depends on low barriers regarding the expertise, computational budget and time required to obtain results and experiment with new techniques. With state-of-the-art approaches relying on task-specific models, multi-GPU setups and weeks of training time, we must find ways to reuse and recombine them to meet these requirements. Instead of designing and training methods for controllable content creation from scratch, we thus present a method to repurpose powerful, existing models for new tasks, even though they have never been designed for them. We formulate this problem as a translation between expert models, which includes common content creation scenarios, such as text-to-image and image-to-image translation, as a special case. As this translation is ambiguous, we learn a generative model of hidden representations of one expert conditioned on hidden representations of the other expert. Working on the level of hidden representations makes optimal use of the computational effort that went into the training of the expert model to produce these efficient, low-dimensional representations. Experiments demonstrate that our approach can translate from BERT, a state-of-the-art expert for text, to BigGAN, a state-of-the-art expert for images, to enable text-to-image generation, which neither of the experts can perform on its own. Additional experiments show the wide applicability of our approach across different conditional image synthesis tasks and improvements over existing methods for image modifications.
26.QEBA: Query-Efficient Boundary-Based Blackbox Attack ⬇️
Machine learning (ML), especially deep neural networks (DNNs) have been widely used in various applications, including several safety-critical ones (e.g. autonomous driving). As a result, recent research about adversarial examples has raised great concerns. Such adversarial attacks can be achieved by adding a small magnitude of perturbation to the input to mislead model prediction. While several whitebox attacks have demonstrated their effectiveness, which assume that the attackers have full access to the machine learning models; blackbox attacks are more realistic in practice. In this paper, we propose a Query-Efficient Boundary-based blackbox Attack (QEBA) based only on model's final prediction labels. We theoretically show why previous boundary-based attack with gradient estimation on the whole gradient space is not efficient in terms of query numbers, and provide optimality analysis for our dimension reduction-based gradient estimation. On the other hand, we conducted extensive experiments on ImageNet and CelebA datasets to evaluate QEBA. We show that compared with the state-of-the-art blackbox attacks, QEBA is able to use a smaller number of queries to achieve a lower magnitude of perturbation with 100% attack success rate. We also show case studies of attacks on real-world APIs including MEGVII Face++ and Microsoft Azure.
27.Heatmap-Based Method for Estimating Drivers' Cognitive Distraction ⬇️
In order to increase road safety, among the visual and manual distractions, modern intelligent vehicles need also to detect cognitive distracted driving (i.e., the drivers mind wandering). In this study, the influence of cognitive processes on the drivers gaze behavior is explored. A novel image-based representation of the driver's eye-gaze dispersion is proposed to estimate cognitive distraction. Data are collected on open highway roads, with a tailored protocol to create cognitive distraction. The visual difference of created shapes shows that a driver explores a wider area in neutral driving compared to distracted driving. Thus, support vector machine (SVM)-based classifiers are trained, and 85.2% of accuracy is achieved for a two-class problem, even with a small dataset. Thus, the proposed method has the discriminative power to recognize cognitive distraction using gaze information. Finally, this work details how this image-based representation could be useful for other cases of distracted driving detection.
28.A Normalized Fully Convolutional Approach to Head and Neck Cancer Outcome Prediction ⬇️
In medical imaging, radiological scans of different modalities serve to enhance different sets of features for clinical diagnosis and treatment planning. This variety enriches the source information that could be used for outcome prediction. Deep learning methods are particularly well-suited for feature extraction from high-dimensional inputs such as images. In this work, we apply a CNN classification network augmented with a FCN preprocessor sub-network to a public TCIA head and neck cancer dataset. The training goal is survival prediction of radiotherapy cases based on pre-treatment FDG PET-CT scans, acquired across 4 different hospitals. We show that the preprocessor sub-network in conjunction with aggregated residual connection leads to improvements over state-of-the-art results when combining both CT and PET input images.
29.Perception-aware time optimal path parameterization for quadrotors ⬇️
The increasing popularity of quadrotors has given rise to a class of predominantly vision-driven vehicles. This paper addresses the problem of perception-aware time optimal path parametrization for quadrotors. Although many different choices of perceptual modalities are available, the low weight and power budgets of quadrotor systems makes a camera ideal for on-board navigation and estimation algorithms. However, this does come with a set of challenges. The limited field of view of the camera can restrict the visibility of salient regions in the environment, which dictates the necessity to consider perception and planning jointly. The main contribution of this paper is an efficient time optimal path parametrization algorithm for quadrotors with limited field of view constraints. We show in a simulation study that a state-of-the-art controller can track planned trajectories, and we validate the proposed algorithm on a quadrotor platform in experiments.
30.Early Screening of SARS-CoV-2 by Intelligent Analysis of X-Ray Images ⬇️
Future SARS-CoV-2 virus outbreak COVID-XX might possibly occur during the next years. However the pathology in humans is so recent that many clinical aspects, like early detection of complications, side effects after recovery or early screening, are currently unknown. In spite of the number of cases of COVID-19, its rapid spread putting many sanitary systems in the edge of collapse has hindered proper collection and analysis of the data related to COVID-19 clinical aspects. We describe an interdisciplinary initiative that integrates clinical research, with image diagnostics and the use of new technologies such as artificial intelligence and radiomics with the aim of clarifying some of SARS-CoV-2 open questions. The whole initiative addresses 3 main points: 1) collection of standardize data including images, clinical data and analytics; 2) COVID-19 screening for its early diagnosis at primary care centers; 3) define radiomic signatures of COVID-19 evolution and associated pathologies for the early treatment of complications. In particular, in this paper we present a general overview of the project, the experimental design and first results of X-ray COVID-19 detection using a classic approach based on HoG and feature selection. Our experiments include a comparison to some recent methods for COVID-19 screening in X-Ray and an exploratory analysis of the feasibility of X-Ray COVID-19 screening. Results show that classic approaches can outperform deep-learning methods in this experimental setting, indicate the feasibility of early COVID-19 screening and that non-COVID infiltration is the group of patients most similar to COVID-19 in terms of radiological description of X-ray. Therefore, an efficient COVID-19 screening should be complemented with other clinical data to better discriminate these cases.
31.Deep Learning for Automatic Pneumonia Detection ⬇️
Pneumonia is the leading cause of death among young children and one of the top mortality causes worldwide. The pneumonia detection is usually performed through examine of chest X-ray radiograph by highly-trained specialists. This process is tedious and often leads to a disagreement between radiologists. Computer-aided diagnosis systems showed the potential for improving diagnostic accuracy. In this work, we develop the computational approach for pneumonia regions detection based on single-shot detectors, squeeze-and-excitation deep convolution neural networks, augmentations and multi-task learning. The proposed approach was evaluated in the context of the Radiological Society of North America Pneumonia Detection Challenge, achieving one of the best results in the challenge.
32.Learning Various Length Dependence by Dual Recurrent Neural Networks ⬇️
Recurrent neural networks (RNNs) are widely used as a memory model for sequence-related problems. Many variants of RNN have been proposed to solve the gradient problems of training RNNs and process long sequences. Although some classical models have been proposed, capturing long-term dependence while responding to short-term changes remains a challenge. To this problem, we propose a new model named Dual Recurrent Neural Networks (DuRNN). The DuRNN consists of two parts to learn the short-term dependence and progressively learn the long-term dependence. The first part is a recurrent neural network with constrained full recurrent connections to deal with short-term dependence in sequence and generate short-term memory. Another part is a recurrent neural network with independent recurrent connections which helps to learn long-term dependence and generate long-term memory. A selection mechanism is added between two parts to help the needed long-term information transfer to the independent neurons. Multiple modules can be stacked to form a multi-layer model for better performance. Our contributions are: 1) a new recurrent model developed based on the divide-and-conquer strategy to learn long and short-term dependence separately, and 2) a selection mechanism to enhance the separating and learning of different temporal scales of dependence. Both theoretical analysis and extensive experiments are conducted to validate the performance of our model, and we also conduct simple visualization experiments and ablation analyses for the model interpretability. Experimental results indicate that the proposed DuRNN model can handle not only very long sequences (over 5000 time steps), but also short sequences very well. Compared with many state-of-the-art RNN models, our model has demonstrated efficient and better performance.
33.A Feature-map Discriminant Perspective for Pruning Deep Neural Networks ⬇️
Network pruning has become the de facto tool to accelerate deep neural networks for mobile and edge applications. Recently, feature-map discriminant based channel pruning has shown promising results, as it aligns well with the CNN objective of differentiating multiple classes and offers better interpretability of the pruning decision. However, existing discriminant-based methods are challenged by computation inefficiency, as there is a lack of theoretical guidance on quantifying the feature-map discriminant power. In this paper, we present a new mathematical formulation to accurately and efficiently quantify the feature-map discriminativeness, which gives rise to a novel criterion,Discriminant Information(DI). We analyze the theoretical property of DI, specifically the non-decreasing property, that makes DI a valid selection criterion. DI-based pruning removes channels with minimum influence to DI value, as they contain little information regarding to the discriminant power. The versatility of DI criterion also enables an intra-layer mixed precision quantization to further compress the network. Moreover, we propose a DI-based greedy pruning algorithm and structure distillation technique to automatically decide the pruned structure that satisfies certain resource budget, which is a common requirement in reality. Extensive experiments demonstratethe effectiveness of our method: our pruned ResNet50 on ImageNet achieves 44% FLOPs reduction without any Top-1 accuracy loss compared to unpruned model
34.Graph-based Proprioceptive Localization Using a Discrete Heading-Length Feature Sequence Matching Approach ⬇️
Proprioceptive localization refers to a new class of robot egocentric localization methods that do not rely on the perception and recognition of external landmarks. These methods are naturally immune to bad weather, poor lighting conditions, or other extreme environmental conditions that may hinder exteroceptive sensors such as a camera or a laser ranger finder. These methods depend on proprioceptive sensors such as inertial measurement units (IMUs) and/or wheel encoders. Assisted by magnetoreception, the sensors can provide a rudimentary estimation of vehicle trajectory which is used to query a prior known map to obtain location. Named as graph-based proprioceptive localization (GBPL), we provide a low cost fallback solution for localization under challenging environmental conditions. As a robot/vehicle travels, we extract a sequence of heading-length values for straight segments from the trajectory and match the sequence with a pre-processed heading-length graph (HLG) abstracted from the prior known map to localize the robot under a graph-matching approach. Using the information from HLG, our location alignment and verification module compensates for trajectory drift, wheel slip, or tire inflation level. We have implemented our algorithm and tested it in both simulated and physical experiments. The algorithm runs successfully in finding robot location continuously and achieves localization accurate at the level that the prior map allows (less than 10m).
35.Towards the Infeasibility of Membership Inference on Deep Models ⬇️
Recent studies propose membership inference (MI) attacks on deep models. Despite the moderate accuracy of such MI attacks, we show that the way the attack accuracy is reported is often misleading and a simple blind attack which is highly unreliable and inefficient in reality can often represent similar accuracy. We show that the current MI attack models can only identify the membership of misclassified samples with mediocre accuracy at best, which only constitute a very small portion of training samples. We analyze several new features that have not been explored for membership inference before, including distance to the decision boundary and gradient norms, and conclude that deep models' responses are mostly indistinguishable among train and non-train samples. Moreover, in contrast with general intuition that deeper models have a capacity to memorize training samples, and, hence, they are more vulnerable to membership inference, we find no evidence to support that and in some cases deeper models are often harder to launch membership inference attack on. Furthermore, despite the common belief, we show that overfitting does not necessarily lead to higher degree of membership leakage. We conduct experiments on MNIST, CIFAR-10, CIFAR-100, and ImageNet, using various model architecture, including LeNet, ResNet, DenseNet, InceptionV3, and Xception. Source code: this https URL}{\color{blue} {this https URL}.
36.An ENAS Based Approach for Constructing Deep Learning Models for Breast Cancer Recognition from Ultrasound Images ⬇️
Deep Convolutional Neural Networks (CNN) provides an "end-to-end" solution for image pattern recognition with impressive performance in many areas of application including medical imaging. Most CNN models of high performance use hand-crafted network architectures that require expertise in CNNs to utilise their potentials. In this paper, we applied the Efficient Neural Architecture Search (ENAS) method to find optimal CNN architectures for classifying breast lesions from ultrasound (US) images. Our empirical study with a dataset of 524 US images shows that the optimal models generated by using ENAS achieve an average accuracy of 89.3%, surpassing other hand-crafted alternatives. Furthermore, the models are simpler in complexity and more efficient. Our study demonstrates that the ENAS approach to CNN model design is a promising direction for classifying ultrasound images of breast lesions.
37.Multiple resolution residual network for automatic thoracic organs-at-risk segmentation from CT ⬇️
We implemented and evaluated a multiple resolution residual network (MRRN) for multiple normal organs-at-risk (OAR) segmentation from computed tomography (CT) images for thoracic radiotherapy treatment (RT) planning. Our approach simultaneously combines feature streams computed at multiple image resolutions and feature levels through residual connections. The feature streams at each level are updated as the images are passed through various feature levels. We trained our approach using 206 thoracic CT scans of lung cancer patients with 35 scans held out for validation to segment the left and right lungs, heart, esophagus, and spinal cord. This approach was tested on 60 CT scans from the open-source AAPM Thoracic Auto-Segmentation Challenge dataset. Performance was measured using the Dice Similarity Coefficient (DSC). Our approach outperformed the best-performing method in the grand challenge for hard-to-segment structures like the esophagus and achieved comparable results for all other structures. Median DSC using our method was 0.97 (interquartile range [IQR]: 0.97-0.98) for the left and right lungs, 0.93 (IQR: 0.93-0.95) for the heart, 0.78 (IQR: 0.76-0.80) for the esophagus, and 0.88 (IQR: 0.86-0.89) for the spinal cord.
38.Segmentation of the Myocardium on Late-Gadolinium Enhanced MRI based on 2.5 D Residual Squeeze and Excitation Deep Learning Model ⬇️
Cardiac left ventricular (LV) segmentation from short-axis MRI acquired 10 minutes after the injection of a contrast agent (LGE-MRI) is a necessary step in the processing allowing the identification and diagnosis of cardiac diseases such as myocardial infarction. However, this segmentation is challenging due to high variability across subjects and the potential lack of contrast between structures. Then, the main objective of this work is to develop an accurate automatic segmentation method based on deep learning models for the myocardial borders on LGE-MRI. To this end, 2.5 D residual neural network integrated with a squeeze and excitation blocks in encoder side with specialized convolutional has been proposed. Late fusion has been used to merge the output of the best trained proposed models from a different set of hyperparameters. A total number of 320 exams (with a mean number of 6 slices per exam) were used for training and 28 exams used for testing. The performance analysis of the proposed ensemble model in the basal and middle slices was similar as compared to intra-observer study and slightly lower at apical slices. The overall Dice score was 82.01% by our proposed method as compared to Dice score of 83.22% obtained from the intra observer study. The proposed model could be used for the automatic segmentation of myocardial border that is a very important step for accurate quantification of no-reflow, myocardial infarction, myocarditis, and hypertrophic cardiomyopathy, among others.
39.Looking back to lower-level information in few-shot learning ⬇️
Humans are capable of learning new concepts from small numbers of examples. In contrast, supervised deep learning models usually lack the ability to extract reliable predictive rules from limited data scenarios when attempting to classify new examples. This challenging scenario is commonly known as few-shot learning. Few-shot learning has garnered increased attention in recent years due to its significance for many real-world problems. Recently, new methods relying on meta-learning paradigms combined with graph-based structures, which model the relationship between examples, have shown promising results on a variety of few-shot classification tasks. However, existing work on few-shot learning is only focused on the feature embeddings produced by the last layer of the neural network. In this work, we propose the utilization of lower-level, supporting information, namely the feature embeddings of the hidden neural network layers, to improve classifier accuracy. Based on a graph-based meta-learning framework, we develop a method called Looking-Back, where such lower-level information is used to construct additional graphs for label propagation in limited data settings. Our experiments on two popular few-shot learning datasets, miniImageNet and tieredImageNet, show that our method can utilize the lower-level information in the network to improve state-of-the-art classification performance.