1.Unsupervised Event-based Learning of Optical Flow, Depth, and Egomotion pdf
In this work, we propose a novel framework for unsupervised learning for event cameras that learns motion information from only the event stream. In particular, we propose an input representation of the events in the form of a discretized volume that maintains the temporal distribution of the events, which we pass through a neural network to predict the motion of the events. This motion is used to attempt to remove any motion blur in the event image. We then propose a loss function applied to the motion compensated event image that measures the motion blur in this image. We train two networks with this framework, one to predict optical flow, and one to predict egomotion and depths, and evaluate these networks on the Multi Vehicle Stereo Event Camera dataset, along with qualitative results from a variety of different scenes.
2.Magnetic Resonance Fingerprinting using Recurrent Neural Networks pdf
Magnetic Resonance Fingerprinting (MRF) is a new approach to quantitative magnetic resonance imaging that allows simultaneous measurement of multiple tissue properties in a single, time-efficient acquisition. Standard MRF reconstructs parametric maps using dictionary matching and lacks scalability due to computational inefficiency. We propose to perform MRF map reconstruction using a recurrent neural network, which exploits the time-dependent information of the MRF signal evolution. We evaluate our method on multiparametric synthetic signals and compare it to existing MRF map reconstruction approaches, including those based on neural networks. Our method achieves state-of-the-art estimates of T1 and T2 values. In addition, the reconstruction time is significantly reduced compared to dictionary-matching based approaches.
3.Generating Diverse and Meaningful Captions pdf
Image Captioning is a task that requires models to acquire a multi-modal understanding of the world and to express this understanding in natural language text. While the state-of-the-art for this task has rapidly improved in terms of n-gram metrics, these models tend to output the same generic captions for similar images. In this work, we address this limitation and train a model that generates more diverse and specific captions through an unsupervised training approach that incorporates a learning signal from an Image Retrieval model. We summarize previous results and improve the state-of-the-art on caption diversity and novelty. We make our source code publicly available online.
4.Very Power Efficient Neural Time-of-Flight pdf
Time-of-Flight (ToF) cameras require active illumination to obtain depth information thus the power of illumination directly affects the performance of ToF cameras. Traditional ToF imaging algorithms is very sensitive to illumination and the depth accuracy degenerates rapidly with the power of it. Therefore, the design of a power efficient ToF camera always creates a painful dilemma for the illumination and the performance trade-off. In this paper, we show that despite the weak signals in many areas under extreme short exposure setting, these signals as a whole can be well utilized through a learning process which directly translates the weak and noisy ToF camera raw to depth map. This creates an opportunity to tackle the aforementioned dilemma and make a very power efficient ToF camera possible. To enable the learning, we collect a comprehensive dataset under a variety of scenes and photographic conditions by a specialized ToF camera. Experiments show that our method is able to robustly process ToF camera raw with the exposure time of one order of magnitude shorter than that used in conventional ToF cameras. In addition to evaluating our approach both quantitatively and qualitatively, we also discuss its implication to designing the next generation power efficient ToF cameras. We will make our dataset and code publicly available.
5.Multi-Shot Sensitivity-Encoded Diffusion MRI using Model-Based Deep Learning (MODL-MUSSELS) pdf
We propose a model-based deep learning architecture for the correction of phase errors in multishot diffusion-weighted echo-planar MRI images. This work is a generalization of MUSSELS, which is a structured low-rank algorithm. We show that an iterative reweighted least-squares implementation of MUSSELS resembles the model-based deep learning (MoDL) framework. We propose to replace the self-learned linear filter bank in MUSSELS with a convolutional neural network, whose parameters are learned from exemplary data. The proposed algorithm reduces the computational complexity of MUSSELS by several orders of magnitude while providing comparable image quality.
6.Balanced Random Forest Classifier in WEKA pdf
Data analysis and machine learning have become an integrative part of the modern scientific methodology, providing automated techniques to predict further information based on observations. One of these classification and regression techniques is the random forest approach. Those decision tree based predictors are best known for their good computational performance and scalability. However, in case of severely imbalanced training data, as often seen in medical studies' data with large control groups, the training algorithm or the sampling process has to be altered in order to improve the prediction quality for minority classes. In this work, a balanced random forest approach for WEKA is proposed. Furthermore, the prediction quality of the unmodified random forest implementation and the new balanced random forest version for WEKA are evaluated against reference implementations in R. Two-class problems on balanced data sets and imbalanced medical studies' data are investigated. A superior prediction quality using the proposed method for imbalanced data is shown compared to the other three techniques.
7.Window detection in aerial texture images of the Berlin 3D CityGML Model pdf
This article explores the usage of the state-of-art neural network Mask R-CNN to be used for window detection of texture files from the CityGML model of Berlin. As texture files are very irregular in terms of size, exposure settings and orientation, we use several parameter optimisation methods to improve the precision. Those textures are cropped from aerial photos, which implies that the angle of the facade, the exposure as well as contrast are calibrated towards the mean and not towards the single facade. The analysis of a single texture image with the human eye itself is challenging: A combination of window and facade estimation and perspective analysis is necessary in order to determine the facades and windows. We train and detect bounding boxes and masks from two data sets with image size 128 and 256. We explore various configuration optimisation methods and the relation of the Region Proposal Network, detected ROIs and the mask output. Our final results shows that the we can improve the average precision scores for both data set sizes, yet the initial AP score varies and leads to different resulting scores.
8.Shallow Cue Guided Deep Visual Tracking via Mixed Models pdf
In this paper, a robust visual tracking approach via mixed model based convolutional neural networks (SDT) is developed. In order to handle abrupt or fast motion, a prior map is generated to facilitate the localization of region of interest (ROI) before the deep tracker is performed. A top-down saliency model with nineteen shallow cues are employed to construct the prior map with online learnt combination weights. Moreover, apart from a holistic deep learner, four local networks are also trained to learn different components of the target. The generated four local heat maps will facilitate to rectify the holistic map by eliminating the distracters to avoid drifting. Furthermore, to guarantee the instance for online update of high quality, a prioritised update strategy is implemented by casting the problem into a label noise problem. The selection probability is designed by considering both confidence values and bio-inspired memory for temporal information integration. Experiments are conducted qualitatively and quantitatively on a set of challenging image sequences. Comparative study demonstrates that the proposed algorithm outperforms other state-of-the-art methods.
9.Multitask Painting Categorization by Deep Multibranch Neural Network pdf
In this work we propose a new deep multibranch neural network to solve the tasks of artist, style, and genre categorization in a multitask formulation. In order to gather clues from low-level texture details and, at the same time, exploit the coarse layout of the painting, the branches of the proposed networks are fed with crops at different resolutions. We propose and compare two different crop strategies: the first one is a random-crop strategy that permits to manage the tradeoff between accuracy and speed; the second one is a smart extractor based on Spatial Transformer Networks trained to extract the most representative subregions. Furthermore, inspired by the results obtained in other domains, we experiment the joint use of hand-crafted features directly computed on the input images along with neural ones. Experiments are performed on a new dataset originally sourced from wikiart.org and hosted by Kaggle, and made suitable for artist, style and genre multitask learning. The dataset here proposed, named MultitaskPainting100k, is composed by 100K paintings, 1508 artists, 125 styles and 41 genres. Our best method, tested on the MultitaskPainting100k dataset, achieves accuracy levels of 56.5%, 57.2%, and 63.6% on the tasks of artist, style and genre prediction respectively.
10.Spatial-Spectral Regularized Local Scaling Cut for Dimensionality Reduction in Hyperspectral Image Classification pdf
Dimensionality reduction (DR) methods have attracted extensive attention to provide discriminative information and reduce the computational burden of the hyperspectral image (HSI) classification. However, the DR methods face many challenges due to limited training samples with high dimensional spectra. To address this issue, a graph-based spatial and spectral regularized local scaling cut (SSRLSC) for DR of HSI data is proposed. The underlying idea of the proposed method is to utilize the information from both the spectral and spatial domains to achieve better classification accuracy than its spectral domain counterpart. In SSRLSC, a guided filter is initially used to smoothen and homogenize the pixels of the HSI data in order to preserve the pixel consistency. This is followed by generation of between-class and within-class dissimilarity matrices in both spectral and spatial domains by regularized local scaling cut (RLSC) and neighboring pixel local scaling cut (NPLSC) respectively. Finally, we obtain the projection matrix by optimizing the updated spatial-spectral between-class and total-class dissimilarity. The effectiveness of the proposed DR algorithm is illustrated with two popular real-world HSI datasets.
11.Learning beamforming in ultrasound imaging pdf
Medical ultrasound (US) is a widespread imaging modality owing its popularity to cost efficiency, portability, speed, and lack of harmful ionizing radiation. In this paper, we demonstrate that replacing the traditional ultrasound processing pipeline with a data-driven, learnable counterpart leads to significant improvement in image quality. Moreover, we demonstrate that greater improvement can be achieved through a learning-based design of the transmitted beam patterns simultaneously with learning an image reconstruction pipeline. We evaluate our method on an in-vivo first-harmonic cardiac ultrasound dataset acquired from volunteers and demonstrate the significance of the learned pipeline and transmit beam patterns on the image quality when compared to standard transmit and receive beamformers used in high frame-rate US imaging. We believe that the presented methodology provides a fundamentally different perspective on the classical problem of ultrasound beam pattern design.
12.Accurate Hand Keypoint Localization on Mobile Devices pdf
We present a novel approach for 2D hand keypoint localization from regular color input. The proposed approach relies on an appropriately designed Convolutional Neural Network (CNN) that computes a set of heatmaps, one per hand keypoint of interest. Extensive experiments with the proposed method compare it against state of the art approaches and demonstrate its accuracy and computational performance on standard, publicly available datasets. The obtained results demonstrate that the proposed method matches or outperforms the competing methods in accuracy, but clearly outperforms them in computational efficiency, making it a suitable building block for applications that require hand keypoint estimation on mobile devices.
13.OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields pdf
Realtime multi-person 2D pose estimation is a key component in enabling machines to have an understanding of people in images and videos. In this work, we present a realtime approach to detect the 2D pose of multiple people in an image. The proposed method uses a nonparametric representation, which we refer to as Part Affinity Fields (PAFs), to learn to associate body parts with individuals in the image. This bottom-up system achieves high accuracy and realtime performance, regardless of the number of people in the image. In previous work, PAFs and body part location estimation were refined simultaneously across training stages. We demonstrate that a PAF-only refinement rather than both PAF and body part location refinement results in a substantial increase in both runtime performance and accuracy. We also present the first combined body and foot keypoint detector, based on an internal annotated foot dataset that we have publicly released. We show that the combined detector not only reduces the inference time compared to running them sequentially, but also maintains the accuracy of each component individually. This work has culminated in the release of OpenPose, the first open-source realtime system for multi-person 2D pose detection, including body, foot, hand, and facial keypoints.
14.Explanatory Graphs for CNNs pdf
This paper introduces a graphical model, namely an explanatory graph, which reveals the knowledge hierarchy hidden inside conv-layers of a pre-trained CNN. Each filter in a conv-layer of a CNN for object classification usually represents a mixture of object parts. We develop a simple yet effective method to disentangle object-part pattern components from each filter. We construct an explanatory graph to organize the mined part patterns, where a node represents a part pattern, and each edge encodes co-activation relationships and spatial relationships between patterns. More crucially, given a pre-trained CNN, the explanatory graph is learned without a need of annotating object parts. Experiments show that each graph node consistently represented the same object part through different images, which boosted the transferability of CNN features. We transferred part patterns in the explanatory graph to the task of part localization, and our method significantly outperformed other approaches.
15.Mining Interpretable AOG Representations from Convolutional Networks via Active Question Answering pdf
In this paper, we present a method to mine object-part patterns from conv-layers of a pre-trained convolutional neural network (CNN). The mined object-part patterns are organized by an And-Or graph (AOG). This interpretable AOG representation consists of a four-layer semantic hierarchy, i.e., semantic parts, part templates, latent patterns, and neural units. The AOG associates each object part with certain neural units in feature maps of conv-layers. The AOG is constructed in a weakly-supervised manner, i.e., very few annotations (e.g., 3-20) of object parts are used to guide the learning of AOGs. We develop a question-answering (QA) method that uses active human-computer communications to mine patterns from a pre-trained CNN, in order to incrementally explain more features in conv-layers. During the learning process, our QA method uses the current AOG for part localization. The QA method actively identifies objects, whose feature maps cannot be explained by the AOG. Then, our method asks people to annotate parts on the unexplained objects, and uses answers to discover CNN patterns corresponding to the newly labeled parts. In this way, our method gradually grows new branches and refines existing branches on the AOG to semanticize CNN representations. In experiments, our method exhibited a high learning efficiency. Our method used about 1/6-1/3 of the part annotations for training, but achieved similar or better part-localization performance than fast-RCNN methods.
16.A Gated Peripheral-Foveal Convolutional Neural Network for Unified Image Aesthetic Prediction pdf
Learning fine-grained details is a key issue in image aesthetic assessment. Most of the previous methods extract the fine-grained details via random cropping strategy, which may undermine the integrity of semantic information. Extensive studies show that humans perceive fine-grained details with a mixture of foveal vision and peripheral vision. Fovea has the highest possible visual acuity and is responsible for seeing the details. The peripheral vision is used for perceiving the broad spatial scene and selecting the attended regions for the fovea. Inspired by these observations, we propose a Gated Peripheral-Foveal Convolutional Neural Network (GPF-CNN). It is a dedicated double-subnet neural network, i.e. a peripheral subnet and a foveal subnet. The former aims to mimic the functions of peripheral vision to encode the holistic information and provide the attended regions. The latter aims to extract fine-grained features on these key regions. Considering that the peripheral vision and foveal vision play different roles in processing different visual stimuli, we further employ a gated information fusion (GIF) network to weight their contributions. The weights are determined through the fully connected layers followed by a sigmoid function. We conduct comprehensive experiments on the standard AVA and Photo.net datasets for unified aesthetic prediction tasks: (i) aesthetic quality classification; (ii) aesthetic score regression; and (iii) aesthetic score distribution prediction. The experimental results demonstrate the effectiveness of the proposed method.
17.Rigid Body Structure and Motion From Two-Frame Point-Correspondences Under Perspective Projection pdf
This paper is concerned with possibility of recovery of motion and structure parameters from multiframes under perspective projection when only points on a rigid body are traced. Free (unrestricted and uncontrolled) pattern of motion between frames is assumed. The major question is how many points and/or how many frames are necessary for the task. It has been shown in an earlier paper {Klopotek:95b} that for orthogonal projection two frames are insufficient for the task. The paper demonstrates that, under perspective projection, that total uncertainty about relative position of focal point versus projection plane makes the recovery of structure and motion from two frames impossible.
18.Dynamic Programming Approach to Template-based OCR pdf
In this paper we propose a dynamic programming solution to the template-based recognition task in OCR case. We formulate a problem of optimal position search for complex objects consisting of parts forming a sequence. We limit the distance between every two adjacent elements with predefined upper and lower thresholds. We choose the sum of penalties for each part in given position as a function to be minimized. We show that such a choice of restrictions allows a faster algorithm to be used than the one for the general form of deformation penalties. We named this algorithm Dynamic Squeezeboxes Packing (DSP) and applied it to solve the two OCR problems: text fields extraction from an image of document Visual Inspection Zone (VIZ) and license plate segmentation. The quality and the performance of resulting solutions were experimentally proved to meet the requirements of the state-of-the-art industrial recognition systems.
19.PnP-AdaNet: Plug-and-Play Adversarial Domain Adaptation Network with a Benchmark at Cross-modality Cardiac Segmentation pdf
Deep convolutional networks have demonstrated the state-of-the-art performance on various medical image computing tasks. Leveraging images from different modalities for the same analysis task holds clinical benefits. However, the generalization capability of deep models on test data with different distributions remain as a major challenge. In this paper, we propose the PnPAdaNet (plug-and-play adversarial domain adaptation network) for adapting segmentation networks between different modalities of medical images, e.g., MRI and CT. We propose to tackle the significant domain shift by aligning the feature spaces of source and target domains in an unsupervised manner. Specifically, a domain adaptation module flexibly replaces the early encoder layers of the source network, and the higher layers are shared between domains. With adversarial learning, we build two discriminators whose inputs are respectively multi-level features and predicted segmentation masks. We have validated our domain adaptation method on cardiac structure segmentation in unpaired MRI and CT. The experimental results with comprehensive ablation studies demonstrate the excellent efficacy of our proposed PnP-AdaNet. Moreover, we introduce a novel benchmark on the cardiac dataset for the task of unsupervised cross-modality domain adaptation. We will make our code and database publicly available, aiming to promote future studies on this challenging yet important research topic in medical imaging.
20.Removing rain streaks by a linear model pdf
Removing rain streaks from a single image continues to draw attentions today in outdoor vision systems. In this paper, we present an efficient method to remove rain streaks. First, the location map of rain pixels needs to be known as precisely as possible, to which we implement a relatively accurate detection of rain streaks by utilizing two characteristics of rain streaks.The key component of our method is to represent the intensity of each detected rain pixel using a linear model:
$p=αs + β$ , where$p$ is the observed intensity of a rain pixel and$s$ represents the intensity of the background (i.e., before rain-affected). To solve$α$ and$β$ for each detected rain pixel, we concentrate on a window centered around it and form an$L_2$ -norm cost function by considering all detected rain pixels within the window, where the corresponding rain-removed intensity of each detected rain pixel is estimated by some neighboring non-rain pixels. By minimizing this cost function, we determine$α$ and$β$ so as to construct the final rain-removed pixel intensity. Compared with several state-of-the-art works, our proposed method can remove rain streaks from a single color image much more efficiently - it offers not only a better visual quality but also a speed-up of several times to one degree of magnitude.
21.Deep Global-Relative Networks for End-to-End 6-DoF Visual Localization and Odometry pdf
For the autonomous navigation of mobile robots, robust and fast visual localization is a challenging task. Although some end-to-end deep neural networks for 6-DoF Visual Odometry (VO) have been reported with promising results, they are still unable to solve the drift problem in long-range navigation. In this paper, we propose the deep global-relative networks (DGRNets), which is a novel global and relative fusion framework based on Recurrent Convolutional Neural Networks (RCNNs). It is designed to jointly estimate global pose and relative localization from consecutive monocular images. DGRNets include feature extraction sub-networks for discriminative feature selection, RCNNs-type relative pose estimation subnetworks for smoothing the VO trajectory and RCNNs-type global pose regression sub-networks for avoiding the accumulation of pose errors. We also propose two loss functions: the first one consists of Cross Transformation Constraints (CTC) that utilize geometric consistency of the adjacent frames to train a more accurate relative sub-networks, and the second one is composed of CTC and Mean Square Error (MSE) between the predicted pose and ground truth used to train the end-to-end DGRNets. The competitive experiments on indoor Microsoft 7-Scenes and outdoor KITTI dataset show that our DGRNets outperform other learning-based monocular VO methods in terms of pose accuracy.
22.Crack Detection Using Enhanced Thresholding on UAV based Collected Images pdf
This paper proposes a thresholding approach for crack detection in an unmanned aerial vehicle (UAV) based infrastructure inspection system. The proposed algorithm performs recursively on the intensity histogram of UAV-taken images to exploit their crack-pixels appearing at the low intensity interval. A quantified criterion of interclass contrast is proposed and employed as an object cost and stop condition for the recursive process. Experiments on different datasets show that our algorithm outperforms different segmentation approaches to accurately extract crack features of some commercial buildings.
23.Physical Attribute Prediction Using Deep Residual Neural Networks pdf
Images taken from the Internet have been used alongside Deep Learning for many different tasks such as: smile detection, ethnicity, hair style, hair colour, gender and age prediction. After witnessing these usages, we were wondering what other attributes can be predicted from facial images available on the Internet. In this paper we tackle the prediction of physical attributes from face images using Convolutional Neural Networks trained on our dataset named FIRW. We crawled around 61, 000 images from the web, then use face detection to crop faces from these real world images. We choose ResNet-50 as our base network architecture. This network was pretrained for the task of face recognition by using the VGG-Face dataset, and we finetune it by using our own dataset to predict physical attributes. Separate networks are trained for the prediction of body type, ethnicity, gender, height and weight; our models achieve the following accuracies for theses tasks, respectively: 84.58%, 87.34%, 97.97%, 70.51%, 63.99%. To validate our choice of ResNet-50 as the base architecture, we also tackle the famous CelebA dataset. Our models achieve an averagy accuracy of 91.19% on CelebA, which is comparable to state-of-the-art approaches.
24.Semi-Supervised Deep Learning for Abnormality Classification in Retinal Images pdf
Supervised deep learning algorithms have enabled significant performance gains in medical image classification tasks. But these methods rely on large labeled datasets that require resource-intensive expert annotation. Semi-supervised generative adversarial network (GAN) approaches offer a means to learn from limited labeled data alongside larger unlabeled datasets, but have not been applied to discern fine-scale, sparse or localized features that define medical abnormalities. To overcome these limitations, we propose a patch-based semi-supervised learning approach and evaluate performance on classification of diabetic retinopathy from funduscopic images. Our semi-supervised approach achieves high AUC with just 10-20 labeled training images, and outperforms the supervised baselines by upto 15% when less than 30% of the training dataset is labeled. Further, our method implicitly enables interpretation of the SSL predictions. As this approach enables good accuracy, resolution and interpretability with lower annotation burden, it sets the pathway for scalable applications of deep learning in clinical imaging.
25.Mini-UAV-based Remote Sensing: Techniques, Applications and Prospectives pdf
The past few decades have witnessed the great progress of unmanned aircraft vehicles (UAVs) in civilian fields, especially in photogrammetry and remote sensing. In contrast with the platforms of manned aircraft and satellite, the UAV platform holds many promising characteristics: flexibility, efficiency, high-spatial/temporal resolution, low cost, easy operation, etc., which make it an effective complement to other remote-sensing platforms and a cost-effective means for remote sensing. Considering the popularity and expansion of UAV-based remote sensing in recent years, this paper provides a systematic survey on the recent advances and future prospectives of UAVs in the remote-sensing community. Specifically, the main challenges and key technologies of remote-sensing data processing based on UAVs are discussed and summarized firstly. Then, we provide an overview of the widespread applications of UAVs in remote sensing. Finally, some prospects for future work are discussed. We hope this paper will provide remote-sensing researchers an overall picture of recent UAV-based remote sensing developments and help guide the further research on this topic.
26.Light Weight Color Image Warping with Inter-Channel Information pdf
Image warping is a necessary step in many multimedia applications such as texture mapping, image-based rendering, panorama stitching, image resizing and optical flow computation etc. Traditionally, color image warping interpolation is performed in each color channel independently. In this paper, we show that the warping quality can be significantly enhanced by exploiting the cross-channel correlation. We design a warping scheme that integrates intra-channel interpolation with cross-channel variation at very low computational cost, which is required for interactive multimedia applications on mobile devices. The effectiveness and efficiency of our method are validated by extensive experiments.
27.Rotation Ensemble Module for Detecting Rotation-Invariant Features pdf
Deep learning has improved many computer vision tasks by utilizing data-driven features instead of using hand-crafted features. However, geometric transformations of input images often degrade the performance of deep learning based methods. In particular, rotation-invariant features are important in computer vision tasks such as face detection, biological feature detection of microscopy images, or robot grasp detection since the input image can be fed into the network with any rotation angle. In this paper, we propose rotation ensemble module (REM) to efficiently train and utilize rotation-invariant features in a deep neural network for computer vision tasks. We evaluated our proposed REM with face detection tasks on FDDB dataset, robotic grasp detection tasks on Cornell dataset, and real robotic grasp tasks with several novel objects. REM based face detection deep neural networks yielded up to 50.8% accuracy in face detection task on FDDB dataset at false rate 20 with IOU 75%, which is about 10.7% higher than the baseline. Robotic grasp detection deep neural networks with our REM also yielded up to 97.6% accuracy in robotic grasp detection on Cornell dataset that is higher than current state-of-the-art performance. In robotic grasp task using a real 4-axis robotic arm with several novel objects, our REM based robotic grasp achieved up to 93.8%, which is significantly higher than the baseline robotic grasps (11.0-56.3%).
28.Learning On-Road Visual Control for Self-Driving Vehicles with Auxiliary Tasks pdf
A safe and robust on-road navigation system is a crucial component of achieving fully automated vehicles. NVIDIA recently proposed an End-to-End algorithm that can directly learn steering commands from raw pixels of a front camera by using one convolutional neural network. In this paper, we leverage auxiliary information aside from raw images and design a novel network structure, called Auxiliary Task Network (ATN), to help boost the driving performance while maintaining the advantage of minimal training data and an End-to-End training method. In this network, we introduce human prior knowledge into vehicle navigation by transferring features from image recognition tasks. Image semantic segmentation is applied as an auxiliary task for navigation. We consider temporal information by introducing an LSTM module and optical flow to the network. Finally, we combine vehicle kinematics with a sensor fusion step. We discuss the benefits of our method over state-of-the-art visual navigation methods both in the Udacity simulation environment and on the real-world Comma.ai dataset.
29.Discriminative analysis of the human cortex using spherical CNNs - a study on Alzheimer's disease diagnosis pdf
In neuroimaging studies, the human cortex is commonly modeled as a sphere to preserve the topological structure of the cortical surface. Cortical neuroimaging measures hence can be modeled in spherical representation. In this work, we explore analyzing the human cortex using spherical CNNs in an Alzheimer's disease (AD) classification task using cortical morphometric measures derived from structural MRI. Our results show superior performance in classifying AD versus cognitively normal and in predicting MCI progression within two years, using structural MRI information only. This work demonstrates for the first time the potential of the spherical CNNs framework in the discriminative analysis of the human cortex and could be extended to other modalities and other neurological diseases.
30.Cross-Database Micro-Expression Recognition: A Benchmark pdf
Cross-database micro-expression recognition (CDMER) is one of recently emerging and interesting problem in micro-expression analysis. CDMER is more challenging than the conventional micro-expression recognition (MER), because the training and testing samples in CDMER come from different micro-expression databases, resulting in the inconsistency of the feature distributions between the training and testing sets. In this paper, we contribute to this topic from three aspects. First, we establish a CDMER experimental evaluation protocol aiming to allow the researchers to conveniently work on this topic and provide a standard platform for evaluating their proposed methods. Second, we conduct benchmark experiments by using NINE state-of-the-art domain adaptation (DA) methods and SIX popular spatiotemporal descriptors for respectively investigating CDMER problem from two different perspectives. Third, we propose a novel DA method called region selective transfer regression (RSTR) to deal with the CDMER task. Our RSTR takes advantage of one important cue for recognizing micro-expressions, i.e., the different contributions of the facial local regions in MER. The overall superior performance of RSTR demonstrates that taking into consideration the important cues benefiting MER, e.g., the facial local region information, contributes to develop effective DA methods for dealing with CDMER problem.
31.Learning Symmetry Consistent Deep CNNs for Face Completion pdf
Deep convolutional networks (CNNs) have achieved great success in face completion to generate plausible facial structures. These methods, however, are limited in maintaining global consistency among face components and recovering fine facial details. On the other hand, reflectional symmetry is a prominent property of face image and benefits face recognition and consistency modeling, yet remaining uninvestigated in deep face completion. In this work, we leverage two kinds of symmetry-enforcing subnets to form a symmetry-consistent CNN model (i.e., SymmFCNet) for effective face completion. For missing pixels on only one of the half-faces, an illumination-reweighted warping subnet is developed to guide the warping and illumination reweighting of the other half-face. As for missing pixels on both of half-faces, we present a generative reconstruction subnet together with a perceptual symmetry loss to enforce symmetry consistency of recovered structures. The SymmFCNet is constructed by stacking generative reconstruction subnet upon illumination-reweighted warping subnet, and can be end-to-end learned from training set of unaligned face images. Experiments show that SymmFCNet can generate high quality results on images with synthetic and real occlusion, and performs favorably against state-of-the-arts.
32.Unsupervised Video Object Segmentation with Distractor-Aware Online Adaptation pdf
Unsupervised video object segmentation is a crucial application in video analysis without knowing any prior information about the objects. It becomes tremendously challenging when multiple objects occur and interact in a given video clip. In this paper, a novel unsupervised video object segmentation approach via distractor-aware online adaptation (DOA) is proposed. DOA models spatial-temporal consistency in video sequences by capturing background dependencies from adjacent frames. Instance proposals are generated by the instance segmentation network for each frame and then selected by motion information as hard negatives if they exist and positives. To adopt high-quality hard negatives, the block matching algorithm is then applied to preceding frames to track the associated hard negatives. General negatives are also introduced in case that there are no hard negatives in the sequence and experiments demonstrate both kinds of negatives (distractors) are complementary. Finally, we conduct DOA using the positive, negative, and hard negative masks to update the foreground/background segmentation. The proposed approach achieves state-of-the-art results on two benchmark datasets, DAVIS 2016 and FBMS-59 datasets.
33.Training on the test set? An analysis of Spampinato et al. [31] pdf
A recent paper [31] claims to classify brain processing evoked in subjects watching ImageNet stimuli as measured with EEG and to use a representation derived from this processing to create a novel object classifier. That paper, together with a series of subsequent papers [8, 15, 17, 20, 21, 30, 35], claims to revolutionize the field by achieving extremely successful results on several computer-vision tasks, including object classification, transfer learning, and generation of images depicting human perception and thought using brain-derived representations measured through EEG. Our novel experiments and analyses demonstrate that their results crucially depend on the block design that they use, where all stimuli of a given class are presented together, and fail with a rapid-event design, where stimuli of different classes are randomly intermixed. The block design leads to classification of arbitrary brain states based on block-level temporal correlations that tend to exist in all EEG data, rather than stimulus-related activity. Because every trial in their test sets comes from the same block as many trials in the corresponding training sets, their block design thus leads to surreptitiously training on the test set. This invalidates all subsequent analyses performed on this data in multiple published papers and calls into question all of the purported results. We further show that a novel object classifier constructed with a random codebook performs as well as or better than a novel object classifier constructed with the representation extracted from EEG data, suggesting that the performance of their classifier constructed with a representation extracted from EEG data does not benefit at all from the brain-derived representation. Our results calibrate the underlying difficulty of the tasks involved and caution against sensational and overly optimistic, but false, claims to the contrary.
34.GD-GAN: Generative Adversarial Networks for Trajectory Prediction and Group Detection in Crowds pdf
This paper presents a novel deep learning framework for human trajectory prediction and detecting social group membership in crowds. We introduce a generative adversarial pipeline which preserves the spatio-temporal structure of the pedestrian's neighbourhood, enabling us to extract relevant attributes describing their social identity. We formulate the group detection task as an unsupervised learning problem, obviating the need for supervised learning of group memberships via hand labeled databases, allowing us to directly employ the proposed framework in different surveillance settings. We evaluate the proposed trajectory prediction and group detection frameworks on multiple public benchmarks, and for both tasks the proposed method demonstrates its capability to better anticipate human sociological behaviour compared to the existing state-of-the-art methods.
35.FML: Face Model Learning from Videos pdf
Monocular image-based 3D reconstruction of faces is a long-standing problem in computer vision. Since image data is a 2D projection of a 3D face, the resulting depth ambiguity makes the problem ill-posed. Most existing methods rely on data-driven priors that are built from limited 3D face scans. In contrast, we propose multi-frame video-based self-supervised training of a deep network that (i) learns a face identity model both in shape and appearance while (ii) jointly learning to reconstruct 3D faces. Our face model is learned using only corpora of in-the-wild video clips collected from the Internet. This virtually endless source of training data enables learning of a highly general 3D face model. In order to achieve this, we propose a novel multi-frame consistency loss that ensures consistent shape and appearance across multiple frames of a subject's face, thus minimizing depth ambiguity. At test time we can use an arbitrary number of frames, so that we can perform both monocular as well as multi-frame reconstruction.
36.Generative One-Shot Learning (GOL): A Semi-Parametric Approach to One-Shot Learning in Autonomous Vision pdf
Highly Autonomous Driving (HAD) systems rely on deep neural networks for the visual perception of the driving environment. Such networks are trained on large manually annotated databases. In this work, a semi-parametric approach to one-shot learning is proposed, with the aim of bypassing the manual annotation step required for training perceptions systems used in autonomous driving. The proposed generative framework, coined Generative One-Shot Learning (GOL), takes as input single one-shot objects, or generic patterns, and a small set of so-called regularization samples used to drive the generative process. New synthetic data is generated as Pareto optimal solutions from one-shot objects using a set of generalization functions built into a generalization generator. GOL has been evaluated on environment perception challenges encountered in autonomous vision.
37.Adam Induces Implicit Weight Sparsity in Rectifier Neural Networks pdf
In recent years, deep neural networks (DNNs) have been applied to various machine leaning tasks, including image recognition, speech recognition, and machine translation. However, large DNN models are needed to achieve state-of-the-art performance, exceeding the capabilities of edge devices. Model reduction is thus needed for practical use. In this paper, we point out that deep learning automatically induces group sparsity of weights, in which all weights connected to an output channel (node) are zero, when training DNNs under the following three conditions: (1) rectified-linear-unit (ReLU) activations, (2) an
$L_2$ -regularized objective function, and (3) the Adam optimizer. Next, we analyze this behavior both theoretically and experimentally, and propose a simple model reduction method: eliminate the zero weights after training the DNN. In experiments on MNIST and CIFAR-10 datasets, we demonstrate the sparsity with various training setups. Finally, we show that our method can efficiently reduce the model size and performs well relative to methods that use a sparsity-inducing regularizer.
38.MID-Fusion: Octree-based Object-Level Multi-Instance Dynamic SLAM pdf
We propose a new multi-instance dynamic RGB-D SLAM system using an object-level octree-based volumetric representation. It can provide robust camera tracking in dynamic environments and at the same time, continuously estimate geometric, semantic, and motion properties for arbitrary objects in the scene. For each incoming frame, we perform instance segmentation to detect objects and refine mask boundaries using geometric and motion information. Meanwhile, we estimate the pose of each existing moving object using an object-oriented tracking method and robustly track the camera pose against the static scene. Based on the estimated camera pose and object poses, we associate segmented masks with existing models and incrementally fuse corresponding colour, depth, semantic, and foreground object probabilities into each object model. In contrast to existing approaches, our system is the first system to generate an object-level dynamic volumetric map from a single RGB-D camera, which can be used directly for robotic tasks. Our method can run at 2-3 Hz on a CPU, excluding the instance segmentation part. We demonstrate its effectiveness by quantitatively and qualitatively testing it on both synthetic and real-world sequences.
39.Fast and Accurate 3D Medical Image Segmentation with Data-swapping Method pdf
Deep neural network models used for medical image segmentation are large because they are trained with high-resolution three-dimensional (3D) images. Graphics processing units (GPUs) are widely used to accelerate the trainings. However, the memory on a GPU is not large enough to train the models. A popular approach to tackling this problem is patch-based method, which divides a large image into small patches and trains the models with these small patches. However, this method would degrade the segmentation quality if a target object spans multiple patches. In this paper, we propose a novel approach for 3D medical image segmentation that utilizes the data-swapping, which swaps out intermediate data from GPU memory to CPU memory to enlarge the effective GPU memory size, for training high-resolution 3D medical images without patching. We carefully tuned parameters in the data-swapping method to obtain the best training performance for 3D U-Net, a widely used deep neural network model for medical image segmentation. We applied our tuning to train 3D U-Net with full-size images of 192 x 192 x 192 voxels in brain tumor dataset. As a result, communication overhead, which is the most important issue, was reduced by 17.1%. Compared with the patch-based method for patches of 128 x 128 x 128 voxels, our training for full-size images achieved improvement on the mean Dice score by 4.48% and 5.32 % for detecting whole tumor sub-region and tumor core sub-region, respectively. The total training time was reduced from 164 hours to 47 hours, resulting in 3.53 times of acceleration.
40.Found in Translation: Learning Robust Joint Representations by Cyclic Translations Between Modalities pdf
Multimodal sentiment analysis is a core research area that studies speaker sentiment expressed from the language, visual, and acoustic modalities. The central challenge in multimodal learning involves inferring joint representations that can process and relate information from these modalities. However, existing work learns joint representations by requiring all modalities as input and as a result, the learned representations may be sensitive to noisy or missing modalities at test time. With the recent success of sequence to sequence (Seq2Seq) models in machine translation, there is an opportunity to explore new ways of learning joint representations that may not require all input modalities at test time. In this paper, we propose a method to learn robust joint representations by translating between modalities. Our method is based on the key insight that translation from a source to a target modality provides a method of learning joint representations using only the source modality as input. We augment modality translations with a cycle consistency loss to ensure that our joint representations retain maximal information from all modalities. Once our translation model is trained with paired multimodal data, we only need data from the source modality at test time for final sentiment prediction. This ensures that our model remains robust from perturbations or missing information in the other modalities. We train our model with a coupled translation-prediction objective and it achieves new state-of-the-art results on multimodal sentiment analysis datasets: CMU-MOSI, ICT-MMMO, and YouTube. Additional experiments show that our model learns increasingly discriminative joint representations with more input modalities while maintaining robustness to missing or perturbed modalities.
41.A Tour of Unsupervised Deep Learning for Medical Image Analysis pdf
Interpretation of medical images for diagnosis and treatment of complex disease from high-dimensional and heterogeneous data remains a key challenge in transforming healthcare. In the last few years, both supervised and unsupervised deep learning achieved promising results in the area of medical imaging and image analysis. Unlike supervised learning which is biased towards how it is being supervised and manual efforts to create class label for the algorithm, unsupervised learning derive insights directly from the data itself, group the data and help to make data driven decisions without any external bias. This review systematically presents various unsupervised models applied to medical image analysis, including autoencoders and its several variants, Restricted Boltzmann machines, Deep belief networks, Deep Boltzmann machine and Generative adversarial network. Future research opportunities and challenges of unsupervised techniques for medical image analysis have also been discussed.
42.Discriminative Supervised Hashing for Cross-Modal similarity Search pdf
With the advantage of low storage cost and high retrieval efficiency, hashing techniques have recently been an emerging topic in cross-modal similarity search. As multiple modal data reflect similar semantic content, many researches aim at learning unified binary codes. However, discriminative hashing features learned by these methods are not adequate. This results in lower accuracy and robustness. We propose a novel hashing learning framework which jointly performs classifier learning, subspace learning and matrix factorization to preserve class-specific semantic content, termed Discriminative Supervised Hashing (DSH), to learn the discrimative unified binary codes for multi-modal data. Besides, reducing the loss of information and preserving the non-linear structure of data, DSH non-linearly projects different modalities into the common space in which the similarity among heterogeneous data points can be measured. Extensive experiments conducted on the three publicly available datasets demonstrate that the framework proposed in this paper outperforms several state-of -the-art methods.