1.AViD Dataset: Anonymized Videos from Diverse Countries ⬇️
We introduce a new public video dataset for action recognition: Anonymized Videos from Diverse countries (AViD). Unlike existing public video datasets, AViD is a collection of action videos from many different countries. The motivation is to create a public dataset that would benefit training and pretraining of action recognition models for everybody, rather than making it useful for limited countries. Further, all the face identities in the AViD videos are properly anonymized to protect their privacy. It also is a static dataset where each video is licensed with the creative commons license. We confirm that most of the existing video datasets are statistically biased to only capture action videos from a limited number of countries. We experimentally illustrate that models trained with such biased datasets do not transfer perfectly to action videos from the other countries, and show that AViD addresses such problem. We also confirm that the new AViD dataset could serve as a good dataset for pretraining the models, performing comparably or better than prior datasets.
2.Scientific Discovery by Generating Counterfactuals using Image Translation ⬇️
Model explanation techniques play a critical role in understanding the source of a model's performance and making its decisions transparent. Here we investigate if explanation techniques can also be used as a mechanism for scientific discovery. We make three contributions: first, we propose a framework to convert predictions from explanation techniques to a mechanism of discovery. Second, we show how generative models in combination with black-box predictors can be used to generate hypotheses (without human priors) that can be critically examined. Third, with these techniques we study classification models for retinal images predicting Diabetic Macular Edema (DME), where recent work showed that a CNN trained on these images is likely learning novel features in the image. We demonstrate that the proposed framework is able to explain the underlying scientific mechanism, thus bridging the gap between the model's performance and human understanding.
3.Camera-Lidar Integration: Probabilistic sensor fusion for semantic mapping ⬇️
An automated vehicle operating in an urban environment must be able to perceive and recognise object/obstacles in a three-dimensional world while navigating in a constantly changing environment. In order to plan and execute accurate sophisticated driving maneuvers, a high-level contextual understanding of the surroundings is essential. Due to the recent progress in image processing, it is now possible to obtain high definition semantic information in 2D from monocular cameras, though cameras cannot reliably provide the highly accurate 3D information provided by lasers. The fusion of these two sensor modalities can overcome the shortcomings of each individual sensor, though there are a number of important challenges that need to be addressed in a probabilistic manner. In this paper, we address the common, yet challenging, lidar/camera/semantic fusion problems which are seldom approached in a wholly probabilistic manner. Our approach is capable of using a multi-sensor platform to build a three-dimensional semantic voxelized map that considers the uncertainty of all of the processes involved. We present a probabilistic pipeline that incorporates uncertainties from the sensor readings (cameras, lidar, IMU and wheel encoders), compensation for the motion of the vehicle, and heuristic label probabilities for the semantic images. We also present a novel and efficient viewpoint validation algorithm to check for occlusions from the camera frames. A probabilistic projection is performed from the camera images to the lidar point cloud. Each labelled lidar scan then feeds into an octree map building algorithm that updates the class probabilities of the map voxels every time a new observation is available. We validate our approach using a set of qualitative and quantitative experimental tests on the USyd Dataset.
4.STaRFlow: A SpatioTemporal Recurrent Cell for Lightweight Multi-Frame Optical Flow Estimation ⬇️
We present a new lightweight CNN-based algorithm for multi-frame optical flow estimation. Our solution introduces a double recurrence over spatial scale and time through repeated use of a generic "STaR" (SpatioTemporal Recurrent) cell. It includes (i) a temporal recurrence based on conveying learned features rather than optical flow estimates; (ii) an occlusion detection process which is coupled with optical flow estimation and therefore uses a very limited number of extra parameters. The resulting STaRFlow algorithm gives state-of-the-art performances on MPI Sintel and Kitti2015 and involves significantly less parameters than all other methods with comparable results.
5.Geometric Style Transfer ⬇️
Neural style transfer (NST), where an input image is rendered in the style of another image, has been a topic of considerable progress in recent years. Research over that time has been dominated by transferring aspects of color and texture, yet these factors are only one component of style. Other factors of style include composition, the projection system used, and the way in which artists warp and bend objects. Our contribution is to introduce a neural architecture that supports transfer of geometric style. Unlike recent work in this area, we are unique in being general in that we are not restricted by semantic content. This new architecture runs prior to a network that transfers texture style, enabling us to transfer texture to a warped image. This form of network supports a second novelty: we extend the NST input paradigm. Users can input content/style pair as is common, or they can chose to input a content/texture-style/geometry-style triple. This three image input paradigm divides style into two parts and so provides significantly greater versatility to the output we can produce. We provide user studies that show the quality of our output, and quantify the importance of geometric style transfer to style recognition by humans.
6.Grading video interviews with fairness considerations ⬇️
There has been considerable interest in predicting human emotions and traits using facial images and videos. Lately, such work has come under criticism for poor labeling practices, inconclusive prediction results and fairness considerations. We present a careful methodology to automatically derive social skills of candidates based on their video response to interview questions. We, for the first time, include video data from multiple countries encompassing multiple ethnicities. Also, the videos were rated by individuals from multiple racial backgrounds, following several best practices, to achieve a consensus and unbiased measure of social skills. We develop two machine-learning models to predict social skills. The first model employs expert-guidance to use plausibly causal features. The second uses deep learning and depends solely on the empirical correlations present in the data. We compare errors of both these models, study the specificity of the models and make recommendations. We further analyze fairness by studying the errors of models by race and gender. We verify the usefulness of our models by determining how well they predict interview outcomes for candidates. Overall, the study provides strong support for using artificial intelligence for video interview scoring, while taking care of fairness and ethical considerations.
7.Weakly Supervised Deep Nuclei Segmentation Using Partial Points Annotation in Histopathology Images ⬇️
Nuclei segmentation is a fundamental task in histopathology image analysis. Typically, such segmentation tasks require significant effort to manually generate accurate pixel-wise annotations for fully supervised training. To alleviate such tedious and manual effort, in this paper we propose a novel weakly supervised segmentation framework based on partial points annotation, i.e., only a small portion of nuclei locations in each image are labeled. The framework consists of two learning stages. In the first stage, we design a semi-supervised strategy to learn a detection model from partially labeled nuclei locations. Specifically, an extended Gaussian mask is designed to train an initial model with partially labeled data. Then, selftraining with background propagation is proposed to make use of the unlabeled regions to boost nuclei detection and suppress false positives. In the second stage, a segmentation model is trained from the detected nuclei locations in a weakly-supervised fashion. Two types of coarse labels with complementary information are derived from the detected points and are then utilized to train a deep neural network. The fully-connected conditional random field loss is utilized in training to further refine the model without introducing extra computational complexity during inference. The proposed method is extensively evaluated on two nuclei segmentation datasets. The experimental results demonstrate that our method can achieve competitive performance compared to the fully supervised counterpart and the state-of-the-art methods while requiring significantly less annotation effort.
8.Impression Space from Deep Template Network ⬇️
It is an innate ability for humans to imagine something only according to their impression, without having to memorize all the details of what they have seen. In this work, we would like to demonstrate that a trained convolutional neural network also has the capability to "remember" its input images. To achieve this, we propose a simple but powerful framework to establish an {\emph{Impression Space}} upon an off-the-shelf pretrained network. This network is referred to as the {\emph{Template Network}} because its filters will be used as templates to reconstruct images from the impression. In our framework, the impression space and image space are bridged by a layer-wise encoding and iterative decoding process. It turns out that the impression space indeed captures the salient features from images, and it can be directly applied to tasks such as unpaired image translation and image synthesis through impression matching without further network training. Furthermore, the impression naturally constructs a high-level common space for different data. Based on this, we propose a mechanism to model the data relations inside the impression space, which is able to reveal the feature similarity between images. Our code will be released.
9.VRUNet: Multi-Task Learning Model for Intent Prediction of Vulnerable Road Users ⬇️
Advanced perception and path planning are at the core for any self-driving vehicle. Autonomous vehicles need to understand the scene and intentions of other road users for safe motion planning. For urban use cases it is very important to perceive and predict the intentions of pedestrians, cyclists, scooters, etc., classified as vulnerable road users (VRU). Intent is a combination of pedestrian activities and long term trajectories defining their future motion. In this paper we propose a multi-task learning model to predict pedestrian actions, crossing intent and forecast their future path from video sequences. We have trained the model on naturalistic driving open-source JAAD dataset, which is rich in behavioral annotations and real world scenarios. Experimental results show state-of-the-art performance on JAAD dataset and how we can benefit from jointly learning and predicting actions and trajectories using 2D human pose features and scene context.
10.Context-Aware Refinement Network Incorporating Structural Connectivity Prior for Brain Midline Delineation ⬇️
Brain midline delineation can facilitate the clinical evaluation of brain midline shift, which plays an important role in the diagnosis and prognosis of various brain pathology. Nevertheless, there are still great challenges with brain midline delineation, such as the largely deformed midline caused by the mass effect and the possible morphological failure that the predicted midline is not a connected curve. To address these challenges, we propose a context-aware refinement network (CAR-Net) to refine and integrate the feature pyramid representation generated by the UNet. Consequently, the proposed CAR-Net explores more discriminative contextual features and a larger receptive field, which is of great importance to predict largely deformed midline. For keeping the structural connectivity of the brain midline, we introduce a novel connectivity regular loss (CRL) to punish the disconnectivity between adjacent coordinates. Moreover, we address the ignored prerequisite of previous regression-based methods that the brain CT image must be in the standard pose. A simple pose rectification network is presented to align the source input image to the standard pose image. Extensive experimental results on the CQ dataset and one inhouse dataset show that the proposed method requires fewer parameters and outperforms three state-of-the-art methods in terms of four evaluation metrics. Code is available at this https URL.
11.Progressive Point Cloud Deconvolution Generation Network ⬇️
In this paper, we propose an effective point cloud generation method, which can generate multi-resolution point clouds of the same shape from a latent vector. Specifically, we develop a novel progressive deconvolution network with the learning-based bilateral interpolation. The learning-based bilateral interpolation is performed in the spatial and feature spaces of point clouds so that local geometric structure information of point clouds can be exploited. Starting from the low-resolution point clouds, with the bilateral interpolation and max-pooling operations, the deconvolution network can progressively output high-resolution local and global feature maps. By concatenating different resolutions of local and global feature maps, we employ the multi-layer perceptron as the generation network to generate multi-resolution point clouds. In order to keep the shapes of different resolutions of point clouds consistent, we propose a shape-preserving adversarial loss to train the point cloud deconvolution generation network. Experimental results demonstrate the effectiveness of our proposed method.
12.Spine Landmark Localization with combining of Heatmap Regression and Direct Coordinate Regression ⬇️
Landmark Localization plays a very important role in processing medical images as well as in disease identification. However, In medical field, it's a challenging task because of the complexity of medical images and the high requirement of accuracy for disease identification and treatment.There are two dominant ways to regress landmark coordination, one using the full convolutional network to regress the heatmaps of landmarks , which is a complex way and heatmap post-process strategies are needed, and the other way is to regress the coordination using CNN + Full Connective Network directly, which is very simple and faster training , but larger dataset and deeper model are needed to achieve higher accuracy. Though with data augmentation and deeper network it can reach a reasonable accuracy, but the accuracy still not reach the requirement of medical field. In addition, a deeper networks also means larger space consumption. To achieve a higher accuracy, we contrived a new landmark regression method which combing heatmap regression and direct coordinate regression base on probability methods and system control theory.
13.Are pathologist-defined labels reproducible? Comparison of the TUPAC16 mitotic figure dataset with an alternative set of labels ⬇️
Pathologist-defined labels are the gold standard for histopathological data sets, regardless of well-known limitations in consistency for some tasks. To date, some datasets on mitotic figures are available and were used for development of promising deep learning-based algorithms. In order to assess robustness of those algorithms and reproducibility of their methods it is necessary to test on several independent datasets. The influence of different labeling methods of these available datasets is currently unknown. To tackle this, we present an alternative set of labels for the images of the auxiliary mitosis dataset of the TUPAC16 challenge. Additional to manual mitotic figure screening, we used a novel, algorithm-aided labeling process, that allowed to minimize the risk of missing rare mitotic figures in the images. All potential mitotic figures were independently assessed by two pathologists. The novel, publicly available set of labels contains 1,999 mitotic figures (+28.80%) and additionally includes 10,483 labels of cells with high similarities to mitotic figures (hard examples). We found significant difference comparing F_1 scores between the original label set (0.549) and the new alternative label set (0.735) using a standard deep learning object detection architecture. The models trained on the alternative set showed higher overall confidence values, suggesting a higher overall label consistency. Findings of the present study show that pathologists-defined labels may vary significantly resulting in notable difference in the model performance. Comparison of deep learning-based algorithms between independent datasets with different labeling methods should be done with caution.
14.DECAPS: Detail-Oriented Capsule Networks ⬇️
Capsule Networks (CapsNets) have demonstrated to be a promising alternative to Convolutional Neural Networks (CNNs). However, they often fall short of state-of-the-art accuracies on large-scale high-dimensional datasets. We propose a Detail-Oriented Capsule Network (DECAPS) that combines the strength of CapsNets with several novel techniques to boost its classification accuracies. First, DECAPS uses an Inverted Dynamic Routing (IDR) mechanism to group lower-level capsules into heads before sending them to higher-level capsules. This strategy enables capsules to selectively attend to small but informative details within the data which may be lost during pooling operations in CNNs. Second, DECAPS employs a Peekaboo training procedure, which encourages the network to focus on fine-grained information through a second-level attention scheme. Finally, the distillation process improves the robustness of DECAPS by averaging over the original and attended image region predictions. We provide extensive experiments on the CheXpert and RSNA Pneumonia datasets to validate the effectiveness of DECAPS. Our networks achieve state-of-the-art accuracies not only in classification (increasing the average area under ROC curves from 87.24% to 92.82% on the CheXpert dataset) but also in the weakly-supervised localization of diseased areas (increasing average precision from 41.7% to 80% for the RSNA Pneumonia detection dataset).
15.Data-Efficient Ranking Distillation for Image Retrieval ⬇️
Recent advances in deep learning has lead to rapid developments in the field of image retrieval. However, the best performing architectures incur significant computational cost. Recent approaches tackle this issue using knowledge distillation to transfer knowledge from a deeper and heavier architecture to a much smaller network. In this paper we address knowledge distillation for metric learning problems. Unlike previous approaches, our proposed method jointly addresses the following constraints i) limited queries to teacher model, ii) black box teacher model with access to the final output representation, and iii) small fraction of original training data without any ground-truth labels. In addition, the distillation method does not require the student and teacher to have same dimensionality. Addressing these constraints reduces computation requirements, dependency on large-scale training datasets and addresses practical scenarios of limited or partial access to private data such as teacher models or the corresponding training data/labels. The key idea is to augment the original training set with additional samples by performing linear interpolation in the final output representation space. Distillation is then performed in the joint space of original and augmented teacher-student sample representations. Results demonstrate that our approach can match baseline models trained with full supervision. In low training sample settings, our approach outperforms the fully supervised approach on two challenging image retrieval datasets, ROxford5k and RParis6k \cite{Roxf} with the least possible teacher supervision.
16.Using Machine Learning to Detect Ghost Images in Automotive Radar ⬇️
Radar sensors are an important part of driver assistance systems and intelligent vehicles due to their robustness against all kinds of adverse conditions, e.g., fog, snow, rain, or even direct sunlight. This robustness is achieved by a substantially larger wavelength compared to light-based sensors such as cameras or lidars. As a side effect, many surfaces act like mirrors at this wavelength, resulting in unwanted ghost detections. In this article, we present a novel approach to detect these ghost objects by applying data-driven machine learning algorithms. For this purpose, we use a large-scale automotive data set with annotated ghost objects. We show that we can use a state-of-the-art automotive radar classifier in order to detect ghost objects alongside real objects. Furthermore, we are able to reduce the amount of false positive detections caused by ghost images in some settings.
17.Continual Adaptation for Deep Stereo ⬇️
Depth estimation from stereo images is carried out with unmatched results by convolutional neural networks trained end-to-end to regress dense disparities. Like for most tasks, this is possible if large amounts of labelled samples are available for training, possibly covering the whole data distribution encountered at deployment time. Being such an assumption systematically met in real applications, the capacity of adapting to any unseen setting becomes of paramount importance. Purposely, we propose a continual adaptation paradigm for deep stereo networks designed to deal with challenging and ever-changing environments. We design a lightweight and modular architecture, Modularly ADaptive Network (MADNet), and formulate Modular ADaptation algorithms(MAD,MAD++) which permit efficient optimization of independent sub-portions of the entire network. In our paradigm the learning signals needed to continuously adapt models online can be sourced from self-supervision via right-to-left image warping or from traditional stereo algorithms. With both sources no other data than the input images being gathered at deployment time are needed.Thus, our network architecture and adaptation algorithms realize the first real-time self-adaptive deep stereo system and pave the way for a new paradigm that can facilitate practical deployment of end-to-end architectures for dense disparity regression.
18.Miss the Point: Targeted Adversarial Attack on Multiple Landmark Detection ⬇️
Recent methods in multiple landmark detection based on deep convolutional neural networks (CNNs) reach high accuracy and improve traditional clinical workflow. However, the vulnerability of CNNs to adversarial-example attacks can be easily exploited to break classification and segmentation tasks. This paper is the first to study how fragile a CNN-based model on multiple landmark detection to adversarial perturbations. Specifically, we propose a novel Adaptive Targeted Iterative FGSM (ATI-FGSM) attack against the state-of-the-art models in multiple landmark detection. The attacker can use ATI-FGSM to precisely control the model predictions of arbitrarily selected landmarks, while keeping other stationary landmarks still, by adding imperceptible perturbations to the original image. A comprehensive evaluation on a public dataset for cephalometric landmark detection demonstrates that the adversarial examples generated by ATI-FGSM break the CNN-based network more effectively and efficiently, compared with the original Iterative FGSM attack. Our work reveals serious threats to patients' health. Furthermore, we discuss the limitations of our method and provide potential defense directions, by investigating the coupling effect of nearby landmarks, i.e., a major source of divergence in our experiments. Our source code is available at this https URL.
19.Distillation Guided Residual Learning for Binary Convolutional Neural Networks ⬇️
It is challenging to bridge the performance gap between Binary CNN (BCNN) and Floating point CNN (FCNN). We observe that, this performance gap leads to substantial residuals between intermediate feature maps of BCNN and FCNN. To minimize the performance gap, we enforce BCNN to produce similar intermediate feature maps with the ones of FCNN. This training strategy, i.e., optimizing each binary convolutional block with block-wise distillation loss derived from FCNN, leads to a more effective optimization to BCNN. It also motivates us to update the binary convolutional block architecture to facilitate the optimization of block-wise distillation loss. Specifically, a lightweight shortcut branch is inserted into each binary convolutional block to complement residuals at each block. Benefited from its Squeeze-and-Interaction (SI) structure, this shortcut branch introduces a fraction of parameters, e.g., 10% overheads, but effectively complements the residuals. Extensive experiments on ImageNet demonstrate the superior performance of our method in both classification efficiency and accuracy, e.g., BCNN trained with our methods achieves the accuracy of 60.45% on ImageNet.
20.Affine Non-negative Collaborative Representation Based Pattern Classification ⬇️
During the past decade, representation-based classification methods have received considerable attention in pattern recognition. In particular, the recently proposed non-negative representation based classification (NRC) method has been reported to achieve promising results in a wide range of classification tasks. However, NRC has two major drawbacks. First, there is no regularization term in the formulation of NRC, which may result in unstable solution and misclassification. Second, NRC ignores the fact that data usually lies in a union of multiple affine subspaces, rather than linear subspaces in practical applications. To address the above issues, this paper presents an affine non-negative collaborative representation (ANCR) model for pattern classification. To be more specific, ANCR imposes a regularization term on the coding vector. Moreover, ANCR introduces an affine constraint to better represent the data from affine subspaces. The experimental results on several benchmarking datasets demonstrate the merits of the proposed ANCR method. The source code of our ANCR is publicly available at this https URL.
21.SeqHAND:RGB-Sequence-Based 3D Hand Pose and Shape Estimation ⬇️
3D hand pose estimation based on RGB images has been studied for a long time. Most of the studies, however, have performed frame-by-frame estimation based on independent static images. In this paper, we attempt to not only consider the appearance of a hand but incorporate the temporal movement information of a hand in motion into the learning framework for better 3D hand pose estimation performance, which leads to the necessity of a large scale dataset with sequential RGB hand images. We propose a novel method that generates a synthetic dataset that mimics natural human hand movements by re-engineering annotations of an extant static hand pose dataset into pose-flows. With the generated dataset, we train a newly proposed recurrent framework, exploiting visuo-temporal features from sequential images of synthetic hands in motion and emphasizing temporal smoothness of estimations with a temporal consistency constraint. Our novel training strategy of detaching the recurrent layer of the framework during domain finetuning from synthetic to real allows preservation of the visuo-temporal features learned from sequential synthetic hand images. Hand poses that are sequentially estimated consequently produce natural and smooth hand movements which lead to more robust estimations. We show that utilizing temporal information for 3D hand pose estimation significantly enhances general pose estimations by outperforming state-of-the-art methods in experiments on hand pose estimation benchmarks.
22.Optical Flow Distillation: Towards Efficient and Stable Video Style Transfer ⬇️
Video style transfer techniques inspire many exciting applications on mobile devices. However, their efficiency and stability are still far from satisfactory. To boost the transfer stability across frames, optical flow is widely adopted, despite its high computational complexity, e.g. occupying over 97% inference time. This paper proposes to learn a lightweight video style transfer network via knowledge distillation paradigm. We adopt two teacher networks, one of which takes optical flow during inference while the other does not. The output difference between these two teacher networks highlights the improvements made by optical flow, which is then adopted to distill the target student network. Furthermore, a low-rank distillation loss is employed to stabilize the output of student network by mimicking the rank of input videos. Extensive experiments demonstrate that our student network without an optical flow module is still able to generate stable video and runs much faster than the teacher network.
23.FC2RN: A Fully Convolutional Corner Refinement Network for Accurate Multi-Oriented Scene Text Detection ⬇️
Recent scene text detection works mainly focus on curve text detection. However, in real applications, the curve texts are more scarce than the multi-oriented ones. Accurate detection of multi-oriented text with large variations of scales, orientations, and aspect ratios is of great significance. Among the multi-oriented detection methods, direct regression for the geometry of scene text shares a simple yet powerful pipeline and gets popular in academic and industrial communities, but it may produce imperfect detections, especially for long texts due to the limitation of the receptive field. In this work, we aim to improve this while keeping the pipeline simple. A fully convolutional corner refinement network (FC2RN) is proposed for accurate multi-oriented text detection, in which an initial corner prediction and a refined corner prediction are obtained at one pass. With a novel quadrilateral RoI convolution operation tailed for multi-oriented scene text, the initial quadrilateral prediction is encoded into the feature maps which can be further used to predict offset between the initial prediction and the ground-truth as well as output a refined confidence score. Experimental results on four public datasets including MSRA-TD500, ICDAR2017-RCTW, ICDAR2015, and COCO-Text demonstrate that FC2RN can outperform the state-of-the-art methods. The ablation study shows the effectiveness of corner refinement and scoring for accurate text localization.
24.$n$-Reference Transfer Learning for Saliency Prediction ⬇️
Benefiting from deep learning research and large-scale datasets, saliency prediction has achieved significant success in the past decade. However, it still remains challenging to predict saliency maps on images in new domains that lack sufficient data for data-hungry models. To solve this problem, we propose a few-shot transfer learning paradigm for saliency prediction, which enables efficient transfer of knowledge learned from the existing large-scale saliency datasets to a target domain with limited labeled examples. Specifically, very few target domain examples are used as the reference to train a model with a source domain dataset such that the training process can converge to a local minimum in favor of the target domain. Then, the learned model is further fine-tuned with the reference. The proposed framework is gradient-based and model-agnostic. We conduct comprehensive experiments and ablation study on various source domain and target domain pairs. The results show that the proposed framework achieves a significant performance improvement. The code is publicly available at \url{this https URL}.
25.Learnable Hollow Kernels for Anatomical Segmentation ⬇️
Segmentation of certain hollow organs, such as the bladder, is especially hard to automate due to their complex geometry, vague intensity gradients in the soft tissues, and a tedious manual process of the data annotation routine. Yet, accurate localization of the walls and the cancer regions in the radiologic images of such organs is an essential step in oncology. To address this issue, we propose a new class of hollow kernels that learn to 'mimic' the contours of the segmented organ, effectively replicating its shape and structural complexity. We train a series of the U-Net-like neural networks using the proposed kernels and demonstrate the superiority of the idea in various spatio-temporal convolution scenarios. Specifically, the dilated hollow-kernel architecture outperforms state-of-the-art spatial segmentation models, whereas the addition of temporal blocks with, e.g., Bi-LSTM, establishes a new multi-class baseline for the bladder segmentation challenge. Our spatio-temporal model based on the hollow kernels reaches the mean dice scores of 0.936, 0.736, and 0.712 for the bladder's inner wall, the outer wall, and the tumor regions, respectively. The results pave the way towards other domain-specific deep learning applications where the shape of the segmented object could be used to form a proper convolution kernel for boosting the segmentation outcome.
26.DCANet: Learning Connected Attentions for Convolutional Neural Networks ⬇️
While self-attention mechanism has shown promising results for many vision tasks, it only considers the current features at a time. We show that such a manner cannot take full advantage of the attention mechanism. In this paper, we present Deep Connected Attention Network (DCANet), a novel design that boosts attention modules in a CNN model without any modification of the internal structure. To achieve this, we interconnect adjacent attention blocks, making information flow among attention blocks possible. With DCANet, all attention blocks in a CNN model are trained jointly, which improves the ability of attention learning. Our DCANet is generic. It is not limited to a specific attention module or base network architecture. Experimental results on ImageNet and MS COCO benchmarks show that DCANet consistently outperforms the state-of-the-art attention modules with a minimal additional computational overhead in all test cases. All code and models are made publicly available.
27.A Benchmark for Inpainting of Clothing Images with Irregular Holes ⬇️
Fashion image understanding is an active research field with a large number of practical applications for the industry. Despite its practical impacts on intelligent fashion analysis systems, clothing image inpainting has not been extensively examined yet. For that matter, we present an extensive benchmark of clothing image inpainting on well-known fashion datasets. Furthermore, we introduce the use of a dilated version of partial convolutions, which efficiently derive the mask update step, and empirically show that the proposed method reduces the required number of layers to form fully-transparent masks. Experiments show that dilated partial convolutions (DPConv) improve the quantitative inpainting performance when compared to the other inpainting strategies, especially it performs better when the mask size is 20% or more of the image. \keywords{image inpainting, fashion image understanding, dilated convolutions, partial convolutions
28.Automatic Detection of Major Freeway Congestion Events Using Wireless Traffic Sensor Data: A Machine Learning Approach ⬇️
Monitoring the dynamics of traffic in major corridors can provide invaluable insight for traffic planning purposes. An important requirement for this monitoring is the availability of methods to automatically detect major traffic events and to annotate the abundance of travel data. This paper introduces a machine learning based approach for reliable detection and characterization of highway traffic congestion events from hundreds of hours of traffic speed data. Indeed, the proposed approach is a generic approach for detection of changes in any given time series, which is the wireless traffic sensor data in the present study. The speed data is initially time-windowed by a ten-hour long sliding window and fed into three Neural Networks that are used to detect the existence and duration of congestion events (slowdowns) in each window. The sliding window captures each slowdown event multiple times and results in increased confidence in congestion detection. The training and parameter tuning are performed on 17,483 hours of data that includes 168 slowdown events. This data is collected and labeled as part of the ongoing probe data validation studies at the Center for Advanced Transportation Technologies (CATT) at the University of Maryland. The Neural networks are carefully trained to reduce the chances of over-fitting to the training data. The experimental results show that this approach is able to successfully detect most of the congestion events, while significantly outperforming a heuristic rule-based approach. Moreover, the proposed approach is shown to be more accurate in estimation of the start-time and end-time of the congestion events.
29.Learning Representations that Support Extrapolation ⬇️
Extrapolation -- the ability to make inferences that go beyond the scope of one's experiences -- is a hallmark of human intelligence. By contrast, the generalization exhibited by contemporary neural network algorithms is largely limited to interpolation between data points in their training corpora. In this paper, we consider the challenge of learning representations that support extrapolation. We introduce a novel visual analogy benchmark that allows the graded evaluation of extrapolation as a function of distance from the convex domain defined by the training data. We also introduce a simple technique, context normalization, that encourages representations that emphasize the relations between objects. We find that this technique enables a significant improvement in the ability to extrapolate, considerably outperforming a number of competitive techniques.
30.Multimodal price prediction ⬇️
Valorization is one of the most heated discussions in the business community, and commodities valorization is one subset in this task. Features of a product is an essential characteristic in valorization and features are categorized into two classes: graphical and non-graphical. Nowadays, the value of products is measured by price. The goal of this research is to achieve an arrangement to predict the price of a product based on specifications of that. We propose five deep learning models to predict the price range of a product, one unimodal and four multimodal systems. The multimodal methods predict based on the image and non-graphical specification of product. As a platform to evaluate the methods, a cellphones dataset has been gathered from GSMArena. In proposed methods, convolutional neural network is an infrastructure. The experimental results show 88.3% F1-score in the best method.
31.A Quick Review on Recent Trends in 3D Point Cloud Data Compression Techniques and the Challenges of Direct Processing in 3D Compressed Domain ⬇️
Automatic processing of 3D Point Cloud data for object detection, tracking and segmentation is the latest trending research in the field of AI and Data Science, which is specifically aimed at solving different challenges of autonomous driving cars and getting real time performance. However, the amount of data that is being produced in the form of 3D point cloud (with LiDAR) is very huge, due to which the researchers are now on the way inventing new data compression algorithms to handle huge volumes of data thus generated. However, compression on one hand has an advantage in overcoming space requirements, but on the other hand, its processing gets expensive due to the decompression, which indents additional computing resources. Therefore, it would be novel to think of developing algorithms that can operate/analyse directly with the compressed data without involving the stages of decompression and recompression (required as many times, the compressed data needs to be operated or analyzed). This research field is termed as Compressed Domain Processing. In this paper, we will quickly review few of the recent state-of-the-art developments in the area of LiDAR generated 3D point cloud data compression, and highlight the future challenges of compressed domain processing of 3D point cloud data.
32.Few Is Enough: Task-Augmented Active Meta-Learning for Brain Cell Classification ⬇️
Deep Neural Networks (or DNNs) must constantly cope with distribution changes in the input data when the task of interest or the data collection protocol changes. Retraining a network from scratch to combat this issue poses a significant cost. Meta-learning aims to deliver an adaptive model that is sensitive to these underlying distribution changes, but requires many tasks during the meta-training process. In this paper, we propose a tAsk-auGmented actIve meta-LEarning (AGILE) method to efficiently adapt DNNs to new tasks by using a small number of training examples. AGILE combines a meta-learning algorithm with a novel task augmentation technique which we use to generate an initial adaptive model. It then uses Bayesian dropout uncertainty estimates to actively select the most difficult samples when updating the model to a new task. This allows AGILE to learn with fewer tasks and a few informative samples, achieving high performance with a limited dataset. We perform our experiments using the brain cell classification task and compare the results to a plain meta-learning model trained from scratch. We show that the proposed task-augmented meta-learning framework can learn to classify new cell types after a single gradient step with a limited number of training samples. We show that active learning with Bayesian uncertainty can further improve the performance when the number of training samples is extremely small. Using only 1% of the training data and a single update step, we achieved 90% accuracy on the new cell type classification task, a 50% points improvement over a state-of-the-art meta-learning algorithm.
33.StyPath: Style-Transfer Data Augmentation For Robust Histology Image Classification ⬇️
The classification of Antibody Mediated Rejection (AMR) in kidney transplant remains challenging even for experienced nephropathologists; this is partly because histological tissue stain analysis is often characterized by low inter-observer agreement and poor reproducibility. One of the implicated causes for inter-observer disagreement is the variability of tissue stain quality between (and within) pathology labs, coupled with the gradual fading of archival sections. Variations in stain colors and intensities can make tissue evaluation difficult for pathologists, ultimately affecting their ability to describe relevant morphological features. Being able to accurately predict the AMR status based on kidney histology images is crucial for improving patient treatment and care. We propose a novel pipeline to build robust deep neural networks for AMR classification based on StyPath, a histological data augmentation technique that leverages a light weight style-transfer algorithm as a means to reduce sample-specific bias. Each image was generated in 1.84 +- 0.03 seconds using a single GTX TITAN V gpu and pytorch, making it faster than other popular histological data augmentation techniques. We evaluated our model using a Monte Carlo (MC) estimate of Bayesian performance and generate an epistemic measure of uncertainty to compare both the baseline and StyPath augmented models. We also generated Grad-CAM representations of the results which were assessed by an experienced nephropathologist; we used this qualitative analysis to elucidate on the assumptions being made by each model. Our results imply that our style-transfer augmentation technique improves histological classification performance (reducing error from 14.8% to 11.5%) and generalization ability.
34.Automatic Detection of COVID-19 Cases on X-ray images Using Convolutional Neural Networks ⬇️
In recent months the world has been surprised by the rapid advance of COVID-19. In order to face this disease and minimize its socio-economic impacts, in addition to surveillance and treatment, diagnosis is a crucial procedure. However, the realization of this is hampered by the delay and the limited access to laboratory tests, demanding new strategies to carry out case triage. In this scenario, deep learning models are being proposed as a possible option to assist the diagnostic process based on chest X-ray and computed tomography images. Therefore, this research aims to automate the process of detecting COVID-19 cases from chest images, using convolutional neural networks (CNN) through deep learning techniques. The results can contribute to expand access to other forms of detection of COVID-19 and to speed up the process of identifying this disease. All databases used, the codes built, and the results obtained from the models' training are available for open access. This action facilitates the involvement of other researchers in enhancing these models since this can contribute to the improvement of results and, consequently, the progress in confronting COVID-19.
35.SIMBA: Specific Identity Markers for Bone Age Assessment ⬇️
Bone Age Assessment (BAA) is a task performed by radiologists to diagnose abnormal growth in a child. In manual approaches, radiologists take into account different identity markers when calculating bone age, i.e., chronological age and gender. However, the current automated Bone Age Assessment methods do not completely exploit the information present in the patient's metadata. With this lack of available methods as motivation, we present SIMBA: Specific Identity Markers for Bone Age Assessment. SIMBA is a novel approach for the task of BAA based on the use of identity markers. For this purpose, we build upon the state-of-the-art model, fusing the information present in the identity markers with the visual features created from the original hand radiograph. We then use this robust representation to estimate the patient's relative bone age: the difference between chronological age and bone age. We validate SIMBA on the Radiological Hand Pose Estimation dataset and find that it outperforms previous state-of-the-art methods. SIMBA sets a trend of a new wave of Computer-aided Diagnosis methods that incorporate all of the data that is available regarding a patient. To promote further research in this area and ensure reproducibility we will provide the source code as well as the pre-trained models of SIMBA.
36.Evaluation of Big Data based CNN Models in Classification of Skin Lesions with Melanoma ⬇️
This chapter presents a methodology for diagnosis of pigmented skin lesions using convolutional neural networks. The architecture is based on convolu-tional neural networks and it is evaluated using new CNN models as well as re-trained modification of pre-existing CNN models were used. The experi-mental results showed that CNN models pre-trained on big datasets for gen-eral purpose image classification when re-trained in order to identify skin le-sion types offer more accurate results when compared to convolutional neural network models trained explicitly from the dermatoscopic images. The best performance was achieved by re-training a modified version of ResNet-50 convolutional neural network with accuracy equal to 93.89%. Analysis on skin lesion pathology type was also performed with classification accuracy for melanoma and basal cell carcinoma being equal to 79.13% and 82.88%, respectively.
37.Joint Blind Deconvolution and Robust Principal Component Analysis for Blood Flow Estimation in Medical Ultrasound Imaging ⬇️
This paper addresses the problem of high-resolution Doppler blood flow estimation from an ultrafast sequence of ultrasound images. Formulating the separation of clutter and blood components as an inverse problem has been shown in the literature to be a good alternative to spatio-temporal singular value decomposition (SVD)-based clutter filtering. In particular, a deconvolution step has recently been embedded in such a problem to mitigate the influence of the experimentally measured point spread function (PSF) of the imaging system. Deconvolution was shown in this context to improve the accuracy of the blood flow reconstruction. However, measuring the PSF requires non-trivial experimental setups. To overcome this limitation, we propose herein a blind deconvolution method able to estimate both the blood component and the PSF from Doppler data. Numerical experiments conducted on simulated and in vivo data demonstrate qualitatively and quantitatively the effectiveness of the proposed approach in comparison with the previous method based on experimentally measured PSF and two other state-of-the-art approaches.
38.Recognition of Instrument-Tissue Interactions in Endoscopic Videos via Action Triplets ⬇️
Recognition of surgical activity is an essential component to develop context-aware decision support for the operating room. In this work, we tackle the recognition of fine-grained activities, modeled as action triplets <instrument, verb, target> representing the tool activity. To this end, we introduce a new laparoscopic dataset, CholecT40, consisting of 40 videos from the public dataset Cholec80 in which all frames have been annotated using 128 triplet classes. Furthermore, we present an approach to recognize these triplets directly from the video data. It relies on a module called Class Activation Guide (CAG), which uses the instrument activation maps to guide the verb and target recognition. To model the recognition of multiple triplets in the same frame, we also propose a trainable 3D Interaction Space, which captures the associations between the triplet components. Finally, we demonstrate the significance of these contributions via several ablation studies and comparisons to baselines on CholecT40.
39.Semi-supervised Task-driven Data Augmentation for Medical Image Segmentation ⬇️
Supervised learning-based segmentation methods typically require a large number of annotated training data to generalize well at test time. In medical applications, curating such datasets is not a favourable option because acquiring a large number of annotated samples from experts is time-consuming and expensive. Consequently, numerous methods have been proposed in the literature for learning with limited annotated examples. Unfortunately, the proposed approaches in the literature have not yet yielded significant gains over random data augmentation for image segmentation, where random augmentations themselves do not yield high accuracy. In this work, we propose a novel task-driven data augmentation method for learning with limited labeled data where the synthetic data generator, is optimized for the segmentation task. The generator of the proposed method models intensity and shape variations using two sets of transformations, as additive intensity transformations and deformation fields. Both transformations are optimized using labeled as well as unlabeled examples in a semi-supervised framework. Our experiments on three medical datasets, namely cardic, prostate and pancreas, show that the proposed approach significantly outperforms standard augmentation and semi-supervised approaches for image segmentation in the limited annotation setting. The code is made publicly available at this https URL_$driven$_$data$_$augmentation.
40.A distance-based loss for smooth and continuous skin layer segmentation in optoacoustic images ⬇️
Raster-scan optoacoustic mesoscopy (RSOM) is a powerful, non-invasive optical imaging technique for functional, anatomical, and molecular skin and tissue analysis. However, both the manual and the automated analysis of such images are challenging, because the RSOM images have very low contrast, poor signal to noise ratio, and systematic overlaps between the absorption spectra of melanin and hemoglobin. Nonetheless, the segmentation of the epidermis layer is a crucial step for many downstream medical and diagnostic tasks, such as vessel segmentation or monitoring of cancer progression. We propose a novel, shape-specific loss function that overcomes discontinuous segmentations and achieves smooth segmentation surfaces while preserving the same volumetric Dice and IoU. Further, we validate our epidermis segmentation through the sensitivity of vessel segmentation. We found a 20
$%$ improvement in Dice for vessel segmentation tasks when the epidermis mask is provided as additional information to the vessel segmentation network.
41.TIMELY: Improving Labeling Consistency in Medical Imaging for Cell Type Classification ⬇️
Diagnosing diseases such as leukemia or anemia requires reliable counts of blood cells. Hematologists usually label and count microscopy images of blood cells manually. In many cases, however, cells in different maturity states are difficult to distinguish, and in combination with image noise and subjectivity, humans are prone to make labeling mistakes. This results in labels that are often not reproducible, which can directly affect the diagnoses. We introduce TIMELY, a probabilistic model that combines pseudotime inference methods with inhomogeneous hidden Markov trees, which addresses this challenge of label inconsistency. We show first on simulation data that TIMELY is able to identify and correct wrong labels with higher precision and recall than baseline methods for labeling correction. We then apply our method to two real-world datasets of blood cell data and show that TIMELY successfully finds inconsistent labels, thereby improving the quality of human-generated labels.
42.Deep Learning-Based Regression and Classification for Automatic Landmark Localization in Medical Images ⬇️
In this study, we propose a fast and accurate method to automatically localize anatomical landmarks in medical images. We employ a global-to-local localization approach using fully convolutional neural networks (FCNNs). First, a global FCNN localizes multiple landmarks through the analysis of image patches, performing regression and classification simultaneously. In regression, displacement vectors pointing from the center of image patches towards landmark locations are determined. In classification, presence of landmarks of interest in the patch is established. Global landmark locations are obtained by averaging the predicted displacement vectors, where the contribution of each displacement vector is weighted by the posterior classification probability of the patch that it is pointing from. Subsequently, for each landmark localized with global localization, local analysis is performed. Specialized FCNNs refine the global landmark locations by analyzing local sub-images in a similar manner, i.e. by performing regression and classification simultaneously and combining the results. Evaluation was performed through localization of 8 anatomical landmarks in CCTA scans, 2 landmarks in olfactory MR scans, and 19 landmarks in cephalometric X-rays. We demonstrate that the method performs similarly to a second observer and is able to localize landmarks in a diverse set of medical images, differing in image modality, image dimensionality, and anatomical coverage.
43.Cross-Attention in Coupled Unmixing Nets for Unsupervised Hyperspectral Super-Resolution ⬇️
The recent advancement of deep learning techniques has made great progress on hyperspectral image super-resolution (HSI-SR). Yet the development of unsupervised deep networks remains challenging for this task. To this end, we propose a novel coupled unmixing network with a cross-attention mechanism, CUCaNet for short, to enhance the spatial resolution of HSI by means of higher-spatial-resolution multispectral image (MSI). Inspired by coupled spectral unmixing, a two-stream convolutional autoencoder framework is taken as backbone to jointly decompose MS and HS data into a spectrally meaningful basis and corresponding coefficients. CUCaNet is capable of adaptively learning spectral and spatial response functions from HS-MS correspondences by enforcing reasonable consistency assumptions on the networks. Moreover, a cross-attention module is devised to yield more effective spatial-spectral information transfer in networks. Extensive experiments are conducted on three widely-used HS-MS datasets in comparison with state-of-the-art HSI-SR models, demonstrating the superiority of the CUCaNet in the HSI-SR application. Furthermore, the codes and datasets will be available at: this https URL.
44.Automatic Segmentation of Non-Tumor Tissues in Glioma MR Brain Images Using Deformable Registration with Partial Convolutional Networks ⬇️
In brain tumor diagnosis and surgical planning, segmentation of tumor regions and accurate analysis of surrounding normal tissues are necessary for physicians. Pathological variability often renders difficulty to register a well-labeled normal atlas to such images and to automatic segment/label surrounding normal brain tissues. In this paper, we propose a new registration approach that first segments brain tumor using a U-Net and then simulates missed normal tissues within the tumor region using a partial convolutional network. Then, a standard normal brain atlas image is registered onto such tumor-removed images in order to segment/label the normal brain tissues. In this way, our new approach greatly reduces the effects of pathological variability in deformable registration and segments the normal tissues surrounding brain tumor well. In experiments, we used MICCAI BraTS2018 T1 tumor images to evaluate the proposed algorithm. By comparing direct registration with the proposed algorithm, the results showed that the Dice coefficient for gray matters was significantly improved for surrounding normal brain tissues.
45.Efficient Unpaired Image Dehazing with Cyclic Perceptual-Depth Supervision ⬇️
Image dehazing without paired haze-free images is of immense importance, as acquiring paired images often entails significant cost. However, we observe that previous unpaired image dehazing approaches tend to suffer from performance degradation near depth borders, where depth tends to vary abruptly. Hence, we propose to anneal the depth border degradation in unpaired image dehazing with cyclic perceptual-depth supervision. Coupled with the dual-path feature re-using backbones of the generators and discriminators, our model achieves
$\mathbf{20.36}$ Peak Signal-to-Noise Ratio (PSNR) on NYU Depth V2 dataset, significantly outperforming its predecessors with reduced Floating Point Operations (FLOPs).
46.OT-driven Multi-Domain Unsupervised Ultrasound Image Artifact Removal using a Single CNN ⬇️
Ultrasound imaging (US) often suffers from distinct image artifacts from various sources. Classic approaches for solving these problems are usually model-based iterative approaches that have been developed specifically for each type of artifact, which are often computationally intensive. Recently, deep learning approaches have been proposed as computationally efficient and high performance alternatives. Unfortunately, in the current deep learning approaches, a dedicated neural network should be trained with matched training data for each specific artifact type. This poses a fundamental limitation in the practical use of deep learning for US, since large number of models should be stored to deal with various US image artifacts. Inspired by the recent success of multi-domain image transfer, here we propose a novel, unsupervised, deep learning approach in which a single neural network can be used to deal with different types of US artifacts simply by changing a mask vector that switches between different target domains. Our algorithm is rigorously derived using an optimal transport (OT) theory for cascaded probability measures. Experimental results using phantom and in vivo data demonstrate that the proposed method can generate high quality image by removing distinct artifacts, which are comparable to those obtained by separately trained multiple neural networks.
47.ROSE: A Retinal OCT-Angiography Vessel Segmentation Dataset and New Model ⬇️
Optical Coherence Tomography Angiography (OCT-A) is a non-invasive imaging technique, and has been increasingly used to image the retinal vasculature at capillary level resolution. However, automated segmentation of retinal vessels in OCT-A has been under-studied due to various challenges such as low capillary visibility and high vessel complexity, despite its significance in understanding many eye-related diseases. In addition, there is no publicly available OCT-A dataset with manually graded vessels for training and validation. To address these issues, for the first time in the field of retinal image analysis we construct a dedicated Retinal OCT-A SEgmentation dataset (ROSE), which consists of 229 OCT-A images with vessel annotations at either centerline-level or pixel level. This dataset has been released for public access to assist researchers in the community in undertaking research in related topics. Secondly, we propose a novel Split-based Coarse-to-Fine vessel segmentation network (SCF-Net), with the ability to detect thick and thin vessels separately. In the SCF-Net, a split-based coarse segmentation (SCS) module is first introduced to produce a preliminary confidence map of vessels, and a split-based refinement (SRN) module is then used to optimize the shape/contour of the retinal microvasculature. Thirdly, we perform a thorough evaluation of the state-of-the-art vessel segmentation models and our SCF-Net on the proposed ROSE dataset. The experimental results demonstrate that our SCF-Net yields better vessel segmentation performance in OCT-A than both traditional methods and other deep learning methods.
48.Hyperspectral Imaging to detect Age, Defects and Individual Nutrient Deficiency in Grapevine Leaves ⬇️
Hyperspectral (HS) imaging was successfully employed in the 380 nm to 1000 nm wavelength range to investigate the efficacy of detecting age, healthiness and individual nutrient deficiency of grapevine leaves collected from vineyards located in central west NSW, Australia. For age detection, the appearance of many healthy grapevine leaves has been examined. Then visually defective leaves were compared with healthy leaves. Control leaves and individual nutrient-deficient leaves (e.g. N, K and Mg) were also analysed. Several features were employed at various stages in the Ultraviolet (UV), Visible (VIS) and Near Infrared (NIR) regions to evaluate the experimental data: mean brightness, mean 1st derivative brightness, variation index, mean spectral ratio, normalised difference vegetation index (NDVI) and standard deviation (SD). Experiment results demonstrate that these features could be utilised with a high degree of effectiveness to compare age, identify unhealthy samples and not only to distinguish from control and nutrient deficiency but also to identify individual nutrient defects. Therefore, our work corroborated that HS imaging has excellent potential as a non-destructive as well as a non-contact method to detect age, healthiness and individual nutrient deficiencies of grapevine leaves
49.Rain Streak Removal in a Video to Improve Visibility by TAWL Algorithm ⬇️
In computer vision applications, the visibility of the video content is crucial to perform analysis for better accuracy. The visibility can be affected by several atmospheric interferences in challenging weather-one of them is the appearance of rain streak. In recent time, rain streak removal achieves lots of interest to the researchers as it has some exciting applications such as autonomous car, intelligent traffic monitoring system, multimedia, etc. In this paper, we propose a novel and simple method by combining three novel extracted features focusing on temporal appearance, wide shape and relative location of the rain streak and we called it TAWL (Temporal Appearance, Width, and Location) method. The proposed TAWL method adaptively uses features from different resolutions and frame rates. Moreover, it progressively processes features from the up-coming frames so that it can remove rain in the real-time. The experiments have been conducted using video sequences with both real rains and synthetic rains to compare the performance of the proposed method against the relevant state-of-the-art methods. The experimental results demonstrate that the proposed method outperforms the state-of-the-art methods by removing more rain streaks while keeping other moving regions.
50.Self-Reflective Variational Autoencoder ⬇️
The Variational Autoencoder (VAE) is a powerful framework for learning probabilistic latent variable generative models. However, typical assumptions on the approximate posterior distribution of the encoder and/or the prior, seriously restrict its capacity for inference and generative modeling. Variational inference based on neural autoregressive models respects the conditional dependencies of the exact posterior, but this flexibility comes at a cost: such models are expensive to train in high-dimensional regimes and can be slow to produce samples. In this work, we introduce an orthogonal solution, which we call self-reflective inference. By redesigning the hierarchical structure of existing VAE architectures, self-reflection ensures that the stochastic flow preserves the factorization of the exact posterior, sequentially updating the latent codes in a recurrent manner consistent with the generative model. We empirically demonstrate the clear advantages of matching the variational posterior to the exact posterior - on binarized MNIST, self-reflective inference achieves state-of-the art performance without resorting to complex, computationally expensive components such as autoregressive layers. Moreover, we design a variational normalizing flow that employs the proposed architecture, yielding predictive benefits compared to its purely generative counterpart. Our proposed modification is quite general and complements the existing literature; self-reflective inference can naturally leverage advances in distribution estimation and generative modeling to improve the capacity of each layer in the hierarchy.
51.Localized Motion Artifact Reduction on Brain MRI Using Deep Learning with Effective Data Augmentation Techniques ⬇️
In-scanner motion degrades the quality of magnetic resonance imaging (MRI) thereby reducing its utility in the detection of clinically relevant abnormalities. We introduce a deep learning-based MRI artifact reduction model (DMAR) to localize and correct head motion artifacts in brain MRI scans. Our approach integrates the latest advances in object detection and noise reduction in Computer Vision. Specifically, DMAR employs a two-stage approach: in the first, degraded regions are detected using the Single Shot Multibox Detector (SSD), and in the second, the artifacts within the found regions are reduced using a convolutional autoencoder (CAE). We further introduce a set of novel data augmentation techniques to address the high dimensionality of MRI images and the scarcity of available data. As a result, our model was trained on a large synthetic dataset of 217,000 images generated from six whole-brain T1-weighted MRI scans obtained from three subjects. DMAR produces convincing visual results when applied to both synthetic test images and 55 real-world motion-affected slices from 18 subjects from the multi-center Autism Brain Imaging Data Exchange study. Quantitatively, depending on the level of degradation, our model achieves a 14.3%-25.6% reduction in RMSE and a 1.38-2.68 dB gain in PSNR on a 5000-sample set of synthetic images. For real-world scans where the ground-truth is unavailable, our model produces a 3.65% reduction in regional standard deviations of image intensity.
52.Improving Adversarial Robustness by Enforcing Local and Global Compactness ⬇️
The fact that deep neural networks are susceptible to crafted perturbations severely impacts the use of deep learning in certain domains of application. Among many developed defense models against such attacks, adversarial training emerges as the most successful method that consistently resists a wide range of attacks. In this work, based on an observation from a previous study that the representations of a clean data example and its adversarial examples become more divergent in higher layers of a deep neural net, we propose the Adversary Divergence Reduction Network which enforces local/global compactness and the clustering assumption over an intermediate layer of a deep neural network. We conduct comprehensive experiments to understand the isolating behavior of each component (i.e., local/global compactness and the clustering assumption) and compare our proposed model with state-of-the-art adversarial training methods. The experimental results demonstrate that augmenting adversarial training with our proposed components can further improve the robustness of the network, leading to higher unperturbed and adversarial predictive performances.
53.Adversarially-learned Inference via an Ensemble of Discrete Undirected Graphical Models ⬇️
Undirected graphical models are compact representations of joint probability distributions over random variables. Given a distribution over inference tasks, graphical models of arbitrary topology can be trained using empirical risk minimization. However, when faced with new task distributions, these models (EGMs) often need to be re-trained. Instead, we propose an inference-agnostic adversarial training framework for producing an ensemble of graphical models (AGMs). The ensemble is optimized to generate data, and inference is learned as a by-product of this endeavor. AGMs perform comparably with EGMs on inference tasks that the latter were specifically optimized for. Most importantly, AGMs show significantly better generalization capabilities across distributions of inference tasks. AGMs are also on par with GibbsNet, a state-of-the-art deep neural architecture, which like AGMs, allows conditioning on any subset of random variables. Finally, AGMs allow fast data sampling, competitive with Gibbs sampling from EGMs.
54.Multi-view Orthonormalized Partial Least Squares: Regularizations and Deep Extensions ⬇️
We establish a family of subspace-based learning method for multi-view learning using the least squares as the fundamental basis. Specifically, we investigate orthonormalized partial least squares (OPLS) and study its important properties for both multivariate regression and classification. Building on the least squares reformulation of OPLS, we propose a unified multi-view learning framework to learn a classifier over a common latent space shared by all views. The regularization technique is further leveraged to unleash the power of the proposed framework by providing three generic types of regularizers on its inherent ingredients including model parameters, decision values and latent projected points. We instantiate a set of regularizers in terms of various priors. The proposed framework with proper choices of regularizers not only can recast existing methods, but also inspire new models. To further improve the performance of the proposed framework on complex real problems, we propose to learn nonlinear transformations parameterized by deep networks. Extensive experiments are conducted to compare various methods on nine data sets with different numbers of views in terms of both feature extraction and cross-modal retrieval.
55.Neural Architecture Search with GBDT ⬇️
Neural architecture search (NAS) with an accuracy predictor that predicts the accuracy of candidate architectures has drawn increasing interests due to its simplicity and effectiveness. Previous works employ neural network based predictors which unfortunately cannot well exploit the tabular data representations of network architectures. As decision tree-based models can better handle tabular data, in this paper, we propose to leverage gradient boosting decision tree (GBDT) as the predictor for NAS and demonstrate that it can improve the prediction accuracy and help to find better architectures than neural network based predictors. Moreover, considering that a better and compact search space can ease the search process, we propose to prune the search space gradually according to important features derived from GBDT using an interpreting tool named SHAP. In this way, NAS can be performed by first pruning the search space (using GBDT as a pruner) and then searching a neural architecture (using GBDT as a predictor), which is more efficient and effective. Experiments on NASBench-101 and ImageNet demonstrate the effectiveness of GBDT for NAS: (1) NAS with GBDT predictor finds top-10 architecture (among all the architectures in the search space) with
$0.18%$ test regret on NASBench-101, and achieves$24.2%$ top-1 error rate on ImageNet; and (2) GBDT based search space pruning and neural architecture search further achieves$23.5%$ top-1 error rate on ImageNet.