Environmental information can provide reliable prior information of human motion intent, which can aid the subject with wearable robotics to walk in complex environments. Previous researches have utilized 1D signal and 2D images to classify environments, but they may face the problems of self-occlusion. Comparatively, 3D point cloud can be more appropriate to depict environments, thus we propose a directional PointNet to classify 3D point cloud directly. By utilizing the orientation information of the point cloud, the directional PointNet can classify daily terrains, including level ground, up stairs, and down stairs, and the classification accuracy achieves 99% for testing set. Moreover, the directional PointNet is more efficient than the previous PointNet because the T-net, which is utilized to estimate the transformation of the point cloud, is removed in this research and the length of the global feature is optimized. The experimental results demonstrate that the directional PointNet can classify the environments robustly and efficiently.
http://arxiv.org/abs/1903.06846
Recognising relevant objects or object states in its environment is a basic capability for an autonomous robot. The dominant approach to object recognition in images and range images is classification by supervised machine learning, nowadays mostly with deep convolutional neural networks (CNNs). This works well for target classes whose variability can be completely covered with training examples. However, a robot moving in the wild, i.e., in an environment that is not known at the time the recognition system is trained, will often face \emph{domain shift}: the training data cannot be assumed to exhaustively cover all the within-class variability that will be encountered in the test data. In that situation, learning is in principle possible, since the training set does capture the defining properties, respectively dissimilarities, of the target classes. But directly training a CNN to predict class probabilities is prone to overfitting to irrelevant correlations between the class labels and the specific subset of the target class that is represented in the training set. We explore the idea to instead learn a Siamese CNN that acts as similarity function between pairs of training examples. Class predictions are then obtained by measuring the similarities between a new test instance and the training samples. We show that the CNN embedding correctly recovers the relative similarities to arbitrary class exemplars in the training set. And that therefore few, randomly picked training exemplars are sufficient to achieve good predictions, making the procedure efficient.
http://arxiv.org/abs/1903.06837
The advent of Generative Adversarial Networks (GANs) has brought about completely novel ways of transforming and manipulating pixels in digital images. GAN based techniques such as Image-to-Image translations, DeepFakes, and other automated methods have become increasingly popular in creating fake images. In this paper, we propose a novel approach to detect GAN generated fake images using a combination of co-occurrence matrices and deep learning. We extract co-occurrence matrices on three color channels in the pixel domain and train a model using a deep convolutional neural network (CNN) framework. Experimental results on two diverse and challenging GAN datasets comprising more than 56,000 images based on unpaired image-to-image translations (cycleGAN [1]) and facial attributes/expressions (StarGAN [2]) show that our approach is promising and achieves more than 99% classification accuracy in both datasets. Further, our approach also generalizes well and achieves good results when trained on one dataset and tested on the other.
http://arxiv.org/abs/1903.06836
In order to operate autonomously, a robot should explore the environment and build a model of each of the surrounding objects. A common approach is to carefully scan the whole workspace. This is time-consuming. It is also often impossible to reach all the viewpoints required to acquire full knowledge about the environment. Humans can perform shape completion of occluded objects by relying on past experience. Therefore, we propose a method that generates images of an object from various viewpoints using a single input RGB image. A deep neural network is trained to imagine the object appearance from many viewpoints. We present the whole pipeline, which takes a single RGB image as input and returns a sequence of RGB and depth images of the object. The method utilizes a CNN-based object detector to extract the object from the natural scene. Then, the proposed network generates a set of RGB and depth images. We show the results both on a synthetic dataset and on real images.
http://arxiv.org/abs/1903.06814
Camera network and multi-camera calibration for external parameters is a necessary step for a variety of contexts in computer vision and robotics, ranging from three-dimensional reconstruction to human activity tracking. This paper describes a method for camera network and/or multi-camera calibration suitable for specific contexts: the cameras may not all have a common field of view, or if they do, there may be some views that are 180 degrees from one another, and the network may be asynchronous. The calibration object required is one or more planar calibration patterns, rigidly attached to one another, and are distinguishable from one another, such as aruco or charuco patterns. We formulate the camera network and/or multi-camera calibration problem in this context using rigidity constraints, represented as a system of equations, and an approximate solution is found through a two-step process. Synthetic and real experiments, including scenarios of a asynchronous camera network and rotating imaging system, demonstrate the method in a variety of settings. Reconstruction accuracy error was less than 0.5 mm for all datasets. This method is suitable for new users to calibrate a camera network, and the modularity of the calibration object also allows for disassembly, shipping, and the use of this method in a variety of large and small spaces.
http://arxiv.org/abs/1903.06811
We propose a random convolutional neural network to generate a feature space in which we study image classification and retrieval performance. Put briefly we apply random convolutional blocks followed by global average pooling to generate a new feature, and we repeat this k times to produce a k-dimensional feature space. This can be interpreted as partitioning the space of image patches with random hyperplanes which we formalize as a random depthwise convolutional neural network. In the network’s final layer we perform image classification and retrieval with the linear support vector machine and k-nearest neighbor classifiers and study other empirical properties. We show that the ratio of image pixel distribution similarity across classes to within classes is higher in our network’s final layer compared to the input space. When we apply the linear support vector machine for image classification we see that the accuracy is higher than if we were to train just the final layer of VGG16, ResNet18, and DenseNet40 with random weights. In the same setting we compare it to an unsupervised feature learning method and find our accuracy to be comparable on CIFAR10 but higher on CIFAR100 and STL10. We see that the accuracy is not far behind that of trained networks, particularly in the top-k setting. For example the top-2 accuracy of our network is near 90% on both CIFAR10 and a 10-class mini ImageNet, and 85% on STL10. We find that k-nearest neighbor gives a comparable precision on the Corel Princeton Image Similarity Benchmark than if we were to use the final layer of trained networks. As with other networks we find that our network fails to a black box attack even though we lack a gradient and use the sign activation. We highlight sensitivity of our network to background as a potential pitfall and an advantage. Overall our work pushes the boundary of what can be achieved with random weights.
http://arxiv.org/abs/1806.05789
We introduce the isoperimetric loss as a regularization criterion for learning the map from a visual representation to a semantic embedding, to be used to transfer knowledge to unknown classes in a zero-shot learning setting. We use a pre-trained deep neural network model as a visual representation of image data, a Word2Vec embedding of class labels, and linear maps between the visual and semantic embedding spaces. However, the spaces themselves are not linear, and we postulate the sample embedding to be populated by noisy samples near otherwise smooth manifolds. We exploit the graph structure defined by the sample points to regularize the estimates of the manifolds by inferring the graph connectivity using a generalization of the isoperimetric inequalities from Riemannian geometry to graphs. Surprisingly, this regularization alone, paired with the simplest baseline model, outperforms the state-of-the-art among fully automated methods in zero-shot learning benchmarks such as AwA and CUB. This improvement is achieved solely by learning the structure of the underlying spaces by imposing regularity.
http://arxiv.org/abs/1903.06781
In this work, we propose a novel system for smart copy-paste, enabling the synthesis of high-quality results given a masked source image content and a target image context as input. Our system naturally resolves both shading and geometric inconsistencies between source and target image, resulting in a merged result image that features the content from the pasted source image, seamlessly pasted into the target context. Our framework is based on a novel training image transformation procedure that allows to train a deep convolutional neural network end-to-end to automatically learn a representation that is suitable for copy-pasting. Our training procedure works with any image dataset without additional information such as labels, and we demonstrate the effectiveness of our system on two popular datasets, high-resolution face images and the more complex Cityscapes dataset. Our technique outperforms the current state of the art on face images, and we show promising results on the Cityscapes dataset, demonstrating that our system generalizes to much higher resolution than the training data.
http://arxiv.org/abs/1903.06763
Overcoming the visual barrier and developing “see-through vision” has been one of mankind’s long-standing dreams. However, visible light cannot travel through opaque obstructions (e.g. walls). Unlike visible light, though, Radio Frequency (RF) signals penetrate many common building objects and reflect highly off humans. This project creates a breakthrough artificial intelligence methodology by which the skeletal structure of a human can be reconstructed with RF even through visual occlusion. In a novel procedural flow, video and RF data are first collected simultaneously using a co-located setup containing an RGB camera and RF antenna array transceiver. Next, the RGB video is processed with a Part Affinity Field computer-vision model to generate ground truth label locations for each keypoint in the human skeleton. Then, a collective deep-learning model consisting of a Residual Convolutional Neural Network, Region Proposal Network, and Recurrent Neural Network 1) extracts spatial features from RF images, 2) detects and crops out all people present in the scene, and 3) aggregates information over dozens of time-steps to piece together the various limbs that reflect signals back to the receiver at different times. A simulator is created to demonstrate the system. This project has impactful applications in medicine, military, search & rescue, and robotics. Especially during a fire emergency, neither visible light nor infrared thermal imaging can penetrate smoke or fire, but RF can. With over 1 million fires reported in the US per year, this technology could save thousands of lives and tens-of-thousands of injuries.
http://arxiv.org/abs/1904.00739
We introduce a large-scale dataset of human actions and eye movements while playing Atari videos games. The dataset currently has 44 hours of gameplay data from 16 games and a total of 2.97 million demonstrated actions. Human subjects played games in a frame-by-frame manner to allow enough decision time in order to obtain near-optimal decisions. This dataset could be potentially used for research in imitation learning, reinforcement learning, and visual saliency.
http://arxiv.org/abs/1903.06754
Bayesian Optimisation (BO) refers to a class of methods for global optimisation of a function $f$ which is only accessible via point evaluations. It is typically used in settings where $f$ is expensive to evaluate. A common use case for BO in machine learning is model selection, where it is not possible to analytically model the generalisation performance of a statistical model, and we resort to noisy and expensive training and validation procedures to choose the best model. Conventional BO methods have focused on Euclidean and categorical domains, which, in the context of model selection, only permits tuning scalar hyper-parameters of machine learning algorithms. However, with the surge of interest in deep learning, there is an increasing demand to tune neural network \emph{architectures}. In this work, we develop NASBOT, a Gaussian process based BO framework for neural architecture search. To accomplish this, we develop a distance metric in the space of neural network architectures which can be computed efficiently via an optimal transport program. This distance might be of independent interest to the deep learning community as it may find applications outside of BO. We demonstrate that NASBOT outperforms other alternatives for architecture search in several cross validation based model selection tasks on multi-layer perceptrons and convolutional neural networks.
https://arxiv.org/abs/1802.07191
Bayesian Optimisation (BO), refers to a suite of techniques for global optimisation of expensive black box functions, which use introspective Bayesian models of the function to efficiently find the optimum. While BO has been applied successfully in many applications, modern optimisation tasks usher in new challenges where conventional methods fail spectacularly. In this work, we present Dragonfly, an open source Python library for scalable and robust BO. Dragonfly incorporates multiple recently developed methods that allow BO to be applied in challenging real world settings; these include better methods for handling higher dimensional domains, methods for handling multi-fidelity evaluations when cheap approximations of an expensive function are available, methods for optimising over structured combinatorial spaces, such as the space of neural network architectures, and methods for handling parallel evaluations. Additionally, we develop new methodological improvements in BO for selecting the Bayesian model, selecting the acquisition function, and optimising over complex domains with different variable types and additional constraints. We compare Dragonfly to a suite of other packages and algorithms for global optimisation and demonstrate that when the above methods are integrated, they enable significant improvements in the performance of BO. The Dragonfly library is available at dragonfly.github.io.
http://arxiv.org/abs/1903.06694
This paper presents a novel tightly-coupled keyframe-based Simultaneous Localization and Mapping (SLAM) system with loop-closing and relocalization capabilities targeted for the underwater domain. Our previous work, SVIn, augmented the state-of-the-art visual-inertial state estimation package OKVIS to accommodate acoustic data from sonar in a non-linear optimization-based framework. This paper addresses drift and loss of localization – one of the main problems affecting other packages in underwater domain – by providing the following main contributions: a robust initialization method to refine scale using depth measurements, a fast preprocessing step to enhance the image quality, and a real-time loop-closing and relocalization method using bag of words (BoW). An additional contribution is the addition of depth measurements from a pressure sensor to the tightly-coupled optimization formulation. Experimental results on datasets collected with a custom-made underwater sensor suite and an autonomous underwater vehicle from challenging underwater environments with poor visibility demonstrate performance never achieved before in terms of accuracy and robustness.
http://arxiv.org/abs/1810.03200
Recent trends have accelerated the development of spatial applications on mobile devices and robots. These include navigation, augmented reality, human-robot interaction, and others. A key enabling technology for such applications is the understanding of the device’s location and the map of the surrounding environment. This generic problem, referred to as Simultaneous Localization and Mapping (SLAM), is an extensively researched topic in robotics. However, visual SLAM algorithms face several challenges including perceptual aliasing and high computational cost. These challenges affect the accuracy, efficiency, and viability of visual SLAM algorithms, especially for long-term SLAM, and their use in resource-constrained mobile devices. A parallel trend is the ubiquity of Wi-Fi routers for quick Internet access in most urban environments. Most robots and mobile devices are equipped with a Wi-Fi radio as well. We propose a method to utilize Wi-Fi received signal strength to alleviate the challenges faced by visual SLAM algorithms. To demonstrate the utility of this idea, this work makes the following contributions: (i) We propose a generic way to integrate Wi-Fi sensing into visual SLAM algorithms, (ii) We integrate such sensing into three well-known SLAM algorithms, (iii) Using four distinct datasets, we demonstrate the performance of such augmentation in comparison to the original visual algorithms and (iv) We compare our work to Wi-Fi augmented FABMAP algorithm. Overall, we show that our approach can improve the accuracy of visual SLAM algorithms by 11% on average and reduce computation time on average by 15% to 25%.
http://arxiv.org/abs/1903.06687
We would like robots to achieve purposeful manipulation by placing any instance from a category of objects into a desired set of goal states. Existing manipulation pipelines typically specify the desired configuration as a target 6-DOF pose and rely on explicitly estimating the pose of the manipulated objects. However, representing an object with a parameterized transformation defined on a fixed template cannot capture large intra-category shape variation, and specifying a target pose at a category level can be physically infeasible or fail to accomplish the task – e.g. knowing the pose and size of a coffee mug relative to some canonical mug is not sufficient to successfully hang it on a rack by its handle. Hence we propose a novel formulation of category-level manipulation that uses semantic 3D keypoints as the object representation. This keypoint representation enables a simple and interpretable specification of the manipulation target as geometric costs and constraints on the keypoints, which flexibly generalizes existing pose-based manipulation methods. Using this formulation, we factor the manipulation policy into instance segmentation, 3D keypoint detection, optimization-based robot action planning and local dense-geometry-based action execution. This factorization allows us to leverage advances in these sub-problems and combine them into a general and effective perception-to-action manipulation pipeline. Our pipeline is robust to large intra-category shape variation and topology changes as the keypoint representation ignores task-irrelevant geometric details. Extensive hardware experiments demonstrate our method can reliably accomplish tasks with never-before seen objects in a category, such as placing shoes and mugs with significant shape variation into category level target configurations.
http://arxiv.org/abs/1903.06684
To move upwind, sailing vessels have to cross the wind by tacking. During this manoeuvre distance made good may be lost and especially smaller vessels may struggle to complete a tack in averse wind and wave conditions. A decision for the best tack manoeuvre needs to be made based on weather and available tack implementations. This paper develops an adaptive probabilistic tack manoeuvre decision method. The order of attempting different tacking strategies is based on previous success within a timeout, combined with an exploration component. This method is successfully demonstrated on the 1m long sailing vessel Black Python. Four strategies for crossing the wind were evaluated through adaptive probabilistic choices, and the best was identified without detailed sensory knowledge of the actual weather conditions. Based on the positive results, further improvements for a better selection process are suggested and the potential of using the collected data to recognise the impact of weather conditions on tacking efforts is recognised.
http://arxiv.org/abs/1903.06677
Research in neural networks in the field of computer vision has achieved remarkable accuracy for point estimation. However, the uncertainty in the estimation is rarely addressed. Uncertainty quantification accompanied by point estimation can lead to a more informed decision, and even improve the prediction quality. In this work, we focus on uncertainty estimation in the domain of crowd counting. We propose a scalable neural network framework with quantification of decomposed uncertainty using a bootstrap ensemble. We demonstrate that the proposed uncertainty quantification method provides additional insight to the crowd counting problem and is simple to implement. We also show that our proposed method outperforms the current state of the art method in many benchmark data sets. To the best of our knowledge, we have the best system for ShanghaiTech part A and B, UCF CC 50, UCSD, and UCF-QNRF datasets.
http://arxiv.org/abs/1903.07427
One of the big attractions of low-dimensional models for gait design has been the ability to compute solutions rapidly, whereas one of their drawbacks has been the difficulty in mapping the solutions back to the target robot. This paper presents a set of tools for rapidly determining solutions for ``humanoids’’ without removing or lumping degrees of freedom. The main tools are (1) C-FROST, an open-source C++ interface for FROST, a direct collocation optimization tool; and (2) multi-threading. The results will be illustrated on a 20-DoF floating-base model for a Cassie-series bipedal robot through numerical calculations and physical experiments.
http://arxiv.org/abs/1807.06614
In this paper we present a deep-learning based framework for direct camera pose regression and refinement using RGB information only. For this aim we introduce a novel framework for camera pose estimation, that regresses the camera pose as well as offers a solely RGB-based solution for camera pose refinement. Utilizing research results of recent camera pose regression methods, we investigate the effect of adversarial networks on convolutional neural networks (CNNs) trained for camera re-localization applications, with the goal to better learn the geometric connection between camera pose and corresponding RGB image. Similar to Generative Adversarial Networks (GANs), in addition to a camera pose regressor, mapping images to poses, we propose to train a discriminator that effectively distinguishes between regressed and ground truth poses. This pose discriminator is conditioned on features extracted from the respective input image to implicitly model the relationship between ground truth or regressed poses, and once learned can be used to update the predicted camera poses and improve the localization accuracy.
http://arxiv.org/abs/1903.06646
Formal methods have provided approaches for investigating software engineering fundamentals and also have high potential to improve current practices in dependability assurance. In this article, we summarise known strengths and weaknesses of formal methods. From the perspective of the assurance of robots and autonomous systems~(RAS), we highlight new opportunities for integrated formal methods and identify threats to their adoption to be mitigated. Based on these opportunities and threats, we develop an agenda for fundamental and empirical research on integrated formal methods and for successful transfer of validated research to RAS assurance. Furthermore, we outline our expectations on useful outcomes of such an agenda.
http://arxiv.org/abs/1812.10103
Adversarial examples — perturbations to the input of a model that elicit large changes in the output — have been shown to be an effective way of assessing the robustness of sequence-to-sequence (seq2seq) models. However, these perturbations only indicate weaknesses in the model if they do not change the input so significantly that it legitimately results in changes in the expected output. This fact has largely been ignored in the evaluations of the growing body of related literature. Using the example of untargeted attacks on machine translation (MT), we propose a new evaluation framework for adversarial attacks on seq2seq models that takes the semantic equivalence of the pre- and post-perturbation input into account. Using this framework, we demonstrate that existing methods may not preserve meaning in general, breaking the aforementioned assumption that source side perturbations should not result in changes in the expected output. We further use this framework to demonstrate that adding additional constraints on attacks allows for adversarial perturbations that are more meaning-preserving, but nonetheless largely change the output sequence. Finally, we show that performing untargeted adversarial training with meaning-preserving attacks is beneficial to the model in terms of adversarial robustness, without hurting test performance. A toolkit implementing our evaluation framework is released at https://github.com/pmichel31415/teapot-nlp.
http://arxiv.org/abs/1903.06620
Various convolutional neural networks (CNNs) were developed recently that achieved accuracy comparable with that of human beings in computer vision tasks such as image recognition, object detection and tracking, etc. Most of these networks, however, process one single frame of image at a time, and may not fully utilize the temporal and contextual correlation typically present in multiple channels of the same image or adjacent frames from a video, thus limiting the achievable throughput. This limitation stems from the fact that existing CNNs operate on deterministic numbers. In this paper, we propose a novel statistical convolutional neural network (SCNN), which extends existing CNN architectures but operates directly on correlated distributions rather than deterministic numbers. By introducing a parameterized canonical model to model correlated data and defining corresponding operations as required for CNN training and inference, we show that SCNN can process multiple frames of correlated images effectively, hence achieving significant speedup over existing CNN models. We use a CNN based video object detection as an example to illustrate the usefulness of the proposed SCNN as a general network model. Experimental results show that even a non-optimized implementation of SCNN can still achieve 178% speedup over existing CNNs with slight accuracy degradation.
http://arxiv.org/abs/1903.07663
Standard 3D reconstruction pipelines assume stationary world, therefore suffer from ghost artifacts'' whenever dynamic objects are present in the scene. Recent approaches has started tackling this issue, however, they typically either only discard dynamic information, represent it using bounding boxes or per-frame depth or rely on approaches that are inherently slow and not suitable to online settings.
We propose an end-to-end system for live reconstruction of large-scale outdoor dynamic environments. We leverage recent advances in computationally efficient data-driven approaches for 6 DoF object pose estimation to segment the scene into objects and stationarybackground’’. This allows us to represent the scene using a time-dependent (dynamic) map, in which each object is explicitly represented as a separate instance and reconstructed in its own volume. For each time step, our dynamic map maintains a relative pose of each volume with respect to the stationary background. Our system operates in incremental manner which is essential for on-line reconstruction, handles large-scale environments with objects at large distances and runs in (near) real-time. We demonstrate the efficacy of our approach on the KITTI dataset, and provide qualitative and quantitative results showing high-quality dense 3D reconstructions of a number of dynamic scenes.
http://arxiv.org/abs/1903.06708
In this paper, we establish a baseline for object reflection symmetry detection in complex backgrounds by presenting a new benchmark and an end-to-end deep learning approach, opening up a promising direction for symmetry detection in the wild. The new benchmark, Sym-PASCAL, spans challenges including object diversity, multi-objects, part-invisibility, and various complex backgrounds that are far beyond those in existing datasets. The end-to-end deep learning approach, referred to as a side-output residual network (SRN), leverages the output residual units (RUs) to fit the errors between the object ground-truth symmetry and the side-outputs of multiple stages. By cascading RUs in a deep-to-shallow manner, SRN exploits the ‘flow’ of errors among multiple stages to address the challenges of fitting complex output with limited convolutional layers, suppressing the complex backgrounds, and effectively matching object symmetry at different scales. SRN is further upgraded to a multi-task side-output residual network (MT-SRN) for joint symmetry and edge detection, demonstrating its generality to image-to-mask learning tasks. Experimental results validate both the challenging aspects of Sym-PASCAL benchmark related to real-world images and the state-of-the-art performance of the proposed SRN approach.
https://arxiv.org/abs/1807.06621
This paper explores the problem of matching entities across different knowledge graphs. Given a query entity in one knowledge graph, we wish to find the corresponding real-world entity in another knowledge graph. We formalize this problem and present two large-scale datasets for this task based on exiting cross-ontology links between DBpedia and Wikidata, focused on several hundred thousand ambiguous entities. Using a classification-based approach, we find that a simple multi-layered perceptron based on representations derived from RDF2Vec graph embeddings of entities in each knowledge graph is sufficient to achieve high accuracy, with only small amounts of training data. The contributions of our work are datasets for examining this problem and strong baselines on which future work can be based.
http://arxiv.org/abs/1903.06607
We propose a new bottom-up method for multi-person 2D human pose estimation that is particularly well suited for urban mobility such as self-driving cars and delivery robots. The new method, PifPaf, uses a Part Intensity Field (PIF) to localize body parts and a Part Association Field (PAF) to associate body parts with each other to form full human poses. Our method outperforms previous methods at low resolution and in crowded, cluttered and occluded scenes thanks to (i) our new composite field PAF encoding fine-grained information and (ii) the choice of Laplace loss for regressions which incorporates a notion of uncertainty. Our architecture is based on a fully convolutional, single-shot, box-free design. We perform on par with the existing state-of-the-art bottom-up method on the standard COCO keypoint task and produce state-of-the-art results on a modified COCO keypoint task for the transportation domain.
http://arxiv.org/abs/1903.06593
Multiagent reinforcement learning algorithms (MARL) have been demonstrated on complex tasks that require the coordination of a team of multiple agents to complete. Existing works have focused on sharing information between agents via centralized critics to stabilize learning or through communication to increase performance, but do not generally look at how information can be shared between agents to address the curse of dimensionality in MARL. We posit that a multiagent problem can be decomposed into a multi-task problem where each agent explores a subset of the state space instead of exploring the entire state space. This paper introduces a multiagent actor-critic algorithm and method for combining knowledge from homogeneous agents through distillation and value-matching that outperforms policy distillation alone and allows further learning in both discrete and continuous action spaces.
http://arxiv.org/abs/1903.06592
In standard Convolutional Neural Networks (CNNs), the receptive fields of artificial neurons in each layer are designed to share the same size. It is well-known in the neuroscience community that the receptive field size of visual cortical neurons are modulated by the stimulus, which has been rarely considered in constructing CNNs. We propose a dynamic selection mechanism in CNNs that allows each neuron to adaptively adjust its receptive field size based on multiple scales of input information. A building block called Selective Kernel (SK) unit is designed, in which multiple branches with different kernel sizes are fused using softmax attention that is guided by the information in these branches. Different attentions on these branches yield different sizes of the effective receptive fields of neurons in the fusion layer. Multiple SK units are stacked to a deep network termed Selective Kernel Networks (SKNets). On the ImageNet and CIFAR benchmarks, we empirically show that SKNet outperforms the existing state-of-the-art architectures with lower model complexity. Detailed analyses show that the neurons in SKNet can capture target objects with different scales, which verifies the capability of neurons for adaptively adjusting their recpeitve field sizes according to the input. The code and models are available at https://github.com/implus/SKNet.
http://arxiv.org/abs/1903.06586
Two-photon transition rates are investigated in resonance to the ground state transition in wurtzite GaN/AlN quantum dots using an 8-band kp-scheme. The ground state transition is two-photon accessible through the electron-hole separation inherent to wurtzite III-nitride heterostructures as a result of internal polarization fields. We show that this built-in parity breaking mechanism can allow deterministic triggering of single-photon emission via coherent two-photon excitation. Radiative lifetimes obtained for single-photon relaxation are in good agreement with available time-resolved micro-photoluminescence experiments, indicating the reliability of the employed computational framework. Two-photon singly-induced emission is explored in terms of possible non-degeneracy and cavity enhancement of two-photon processes.
https://arxiv.org/abs/1903.06575
In this paper, we introduce a new problem of manipulating a given video by inserting other videos into it. Our main task is, given an object video and a scene video, to insert the object video at a user-specified location in the scene video so that the resulting video looks realistic. We aim to handle different object motions and complex backgrounds without expensive segmentation annotations. As it is difficult to collect training pairs for this problem, we synthesize fake training pairs that can provide helpful supervisory signals when training a neural network with unpaired real data. The proposed network architecture can take both real and fake pairs as input and perform both supervised and unsupervised training in an adversarial learning scheme. To synthesize a realistic video, the network renders each frame based on the current input and previous frames. Within this framework, we observe that injecting noise into previous frames while generating the current frame stabilizes training. We conduct experiments on real-world videos in object tracking and person re-identification benchmark datasets. Experimental results demonstrate that the proposed algorithm is able to synthesize long sequences of realistic videos with a given object video inserted.
http://arxiv.org/abs/1903.06571
Most digital camera pipelines use color constancy methods to reduce the influence of illumination and camera sensor on the colors of scene objects. The highest accuracy of color correction is obtained with learning-based color constancy methods, but they require a significant amount of calibrated training images with known ground-truth illumination. Such calibration is time consuming, preferably done for each sensor individually, and therefore a major bottleneck in acquiring high color constancy accuracy. Statistics-based methods do not require calibrated training images, but they are less accurate. In this paper an unsupervised learning-based method is proposed that learns its parameter values after approximating the unknown ground-truth illumination of the training images, thus avoiding calibration. In terms of accuracy the proposed method outperforms all statistics-based and many learning-based methods. An extension of the method is also proposed, which learns the needed parameters from non-calibrated images taken with one sensor and which can then be successfully applied to images taken with another sensor. This effectively enables inter-camera unsupervised learning for color constancy. Additionally, a new high quality color constancy benchmark dataset with 1707 calibrated images is created, used for testing, and made publicly available. The results are presented and discussed. The source code and the dataset are available at this http URL
http://arxiv.org/abs/1712.00436
Different empirical models have been developed for cloud detection. There is a growing interest in using the ground-based sky/cloud images for this purpose. Several methods exist that perform binary segmentation of clouds. In this paper, we propose to use a deep learning architecture (U-Net) to perform multi-label sky/cloud image segmentation. The proposed approach outperforms recent literature by a large margin.
http://arxiv.org/abs/1903.06562
We present EAT2seq: a novel method to architect automatic linguistic transformations for a number of tasks, including controlled grammatical or lexical changes, style transfer, text generation, and machine translation. Our approach consists in creating an abstract representation of a sentence’s meaning and grammar, which we use as input to an encoder-decoder network trained to reproduce the original sentence. Manipulating the abstract representation allows the transformation of sentences according to user-provided parameters, both grammatically and lexically, in any combination. The same architecture can further be used for controlled text generation, and has additional promise for machine translation. This strategy holds the promise of enabling many tasks that were hitherto outside the scope of NLP techniques for want of sufficient training data. We provide empirical evidence for the effectiveness of our approach by reproducing and transforming English sentences, and evaluating the results both manually and automatically. A single model trained on monolingual data is used for all tasks without any task-specific training. For a model trained on 8.5 million sentences, we report a BLEU score of 74.45 for reproduction, and scores between 55.29 and 81.82 for back-and-forth grammatical transformations across 14 category pairs.
http://arxiv.org/abs/1902.09381
The golf swing is a complex movement requiring considerable full-body coordination to execute proficiently. As such, it is the subject of frequent scrutiny and extensive biomechanical analyses. In this paper, we introduce the notion of golf swing sequencing for detecting key events in the golf swing and facilitating golf swing analysis. To enable consistent evaluation of golf swing sequencing performance, we also introduce the benchmark database GolfDB, consisting of 1400 high-quality golf swing videos, each labeled with event frames, bounding box, player name and sex, club type, and view type. Furthermore, to act as a reference baseline for evaluating golf swing sequencing performance on GolfDB, we propose a lightweight deep neural network called SwingNet, which possesses a hybrid deep convolutional and recurrent neural network architecture. SwingNet correctly detects eight golf swing events at an average rate of 76.1%, and six out of eight events at a rate of 91.8%. In line with the proposed baseline SwingNet, we advocate the use of computationally efficient models in future research to promote in-the-field analysis via deployment on readily-available mobile devices.
http://arxiv.org/abs/1903.06528
While deep learning has seen many recent applications to drug discovery, most have focused on predicting activity or toxicity directly from chemical structure. Phenotypic changes exhibited in cellular images are also indications of the mechanism of action (MoA) of chemical compounds. In this paper, we show how pre-trained convolutional image features can be used to assist scientists in discovering interesting chemical clusters for further investigation. Our method reduces the dimensionality of raw fluorescent stained images from a high throughput imaging (HTI) screen, producing an embedding space that groups together images with similar cellular phenotypes. Running standard unsupervised clustering on this embedding space yields a set of distinct phenotypic clusters. This allows scientists to further select and focus on interesting clusters for downstream analyses. We validate the consistency of our embedding space qualitatively with t-sne visualizations, and quantitatively by measuring embedding variance among images that are known to be similar. Results suggested the usefulness of our proposed workflow using deep learning and clustering and it can lead to robust HTI screening and compound triage.
http://arxiv.org/abs/1903.06516
We provide a comprehensive investigation of different custom and off-the-shelf architectures as well as different approaches to generating feature vectors for offensive language detection. We also show that these approaches work well on small and noisy datasets such as on the Offensive Language Identification Dataset (OLID), so it should be possible to use them for other applications.
http://arxiv.org/abs/1903.07445
We tackle the problem of finding good architectures for multimodal classification problems. We propose a novel and generic search space that spans a large number of possible fusion architectures. In order to find an optimal architecture for a given dataset in the proposed search space, we leverage an efficient sequential model-based exploration approach that is tailored for the problem. We demonstrate the value of posing multimodal fusion as a neural architecture search problem by extensive experimentation on a toy dataset and two other real multimodal datasets. We discover fusion architectures that exhibit state-of-the-art performance for problems with different domain and dataset size, including the NTU RGB+D dataset, the largest multi-modal action recognition dataset available.
http://arxiv.org/abs/1903.06496
Syntactic analysis plays an important role in semantic parsing, but the nature of this role remains a topic of ongoing debate. The debate has been constrained by the scarcity of empirical comparative studies between syntactic and semantic schemes, which hinders the development of parsing methods informed by the details of target schemes and constructions. We target this gap, and take Universal Dependencies (UD) and UCCA as a test case. After abstracting away from differences of convention or formalism, we find that most content divergences can be ascribed to: (1) UCCA’s distinction between a Scene and a non-Scene; (2) UCCA’s distinction between primary relations, secondary ones and participants; (3) different treatment of multi-word expressions, and (4) different treatment of inter-clause linkage. We further discuss the long tail of cases where the two schemes take markedly different approaches. Finally, we show that the proposed comparison methodology can be used for fine-grained evaluation of UCCA parsing, highlighting both challenges and potential sources for improvement. The substantial differences between the schemes suggest that semantic parsers are likely to benefit downstream text understanding applications beyond their syntactic counterparts.
http://arxiv.org/abs/1903.06494
With the development of technology, the usage areas and importance of biometric systems have started to increase. Since the characteristics of each person are different from each other, a single model biometric system can yield successful results. However, because the characteristics of twin people are very close to each other, multiple biometric systems including multiple characteristics of individuals will be more appropriate and will increase the recognition rate. In this study, a multiple biometric recognition system consisting of a combination of multiple algorithms and multiple models was developed to distinguish people from other people and their twins. Ear and voice biometric data were used for the multimodal model and 38 pair of twin ear images and sound recordings were used in the data set. Sound and ear recognition rates were obtained using classical (hand-crafted) and deep learning algorithms. The results obtained were combined with the score level fusion method to achieve a success rate of 94.74% in rank-1 and 100% in rank -2.
http://arxiv.org/abs/1903.07981
This paper introduces the `Projectron’ as a new neural network architecture that uses Radon projections to both classify and represent medical images. The motivation is to build shallow networks which are more interpretable in the medical imaging domain. Radon transform is an established technique that can reconstruct images from parallel projections. The Projectron first applies global Radon transform to each image using equidistant angles and then feeds these transformations for encoding to a single layer of neurons followed by a layer of suitable kernels to facilitate a linear separation of projections. Finally, the Projectron provides the output of the encoding as an input to two more layers for final classification. We validate the Projectron on five publicly available datasets, a general dataset (namely MNIST) and four medical datasets (namely Emphysema, IDC, IRMA, and Pneumonia). The results are encouraging as we compared the Projectron’s performance against MLPs with raw images and Radon projections as inputs, respectively. Experiments clearly demonstrate the potential of the proposed Projectron for representing/classifying medical images.
http://arxiv.org/abs/1904.00740
Systems which incrementally create 3D semantic maps from image sequences must store and update representations of both geometry and semantic entities. However, while there has been much work on the correct formulation for geometrical estimation, state-of-the-art systems usually rely on simple semantic representations which store and update independent label estimates for each surface element (depth pixels, surfels, or voxels). Spatial correlation is discarded, and fused label maps are incoherent and noisy. We introduce a new compact and optimisable semantic representation by training a variational auto-encoder that is conditioned on a colour image. Using this learned latent space, we can tackle semantic label fusion by jointly optimising the low-dimenional codes associated with each of a set of overlapping images, producing consistent fused label maps which preserve spatial correlation. We also show how this approach can be used within a monocular keyframe based semantic mapping system where a similar code approach is used for geometry. The probabilistic formulation allows a flexible formulation where we can jointly estimate motion, geometry and semantics in a unified optimisation.
http://arxiv.org/abs/1903.06482
We propose DeepHuman, a deep learning based framework for 3D human reconstruction from a single RGB image. Since this problem is highly intractable, we adopt a stage-wise, coarse-to-fine method consisting of three steps, namely body shape/pose estimation, surface reconstruction and frontal surface detail refinement. Once a body is estimated from the given image, our method generates a dense semantic representation from it. The representations not only encodes body shape and pose but also bridges the 2D image plane and 3D space. An image-guided volume-to-volume translation CNN is introduced to reconstruct the surface given the input image and the dense semantic representation. One key feature of our network is that it fuses different scales of image features into the 3D space through volumetric feature transformation, which helps to recover details of the subject’s surface geometry. The details on the frontal areas of the surface are further refined through a normal map refinement network. The normal refinement network can be concatenated with the volume generation network using our proposed volumetric normal projection layer. We also contribute THuman, a 3D real-world human model dataset containing approximately 7000 models. The whole network is trained using training data generated from the dataset. Overall, due to the specific design of our network and the diversity in our dataset, our method enables 3D human reconstruction given only a single image and outperforms state-of-the-art approaches.
http://arxiv.org/abs/1903.06473
With the tremendous growth in the number of scientific papers being published, searching for references while writing a scientific paper is a time-consuming process. A technique that could add a reference citation at the appropriate place in a sentence will be beneficial. In this perspective, context-aware citation recommendation has been researched upon for around two decades. Many researchers have utilized the text data called the context sentence, which surrounds the citation tag, and the metadata of the target paper to find the appropriate cited research. However, the lack of well-organized benchmarking datasets and no model that can attain high performance has made the research difficult. In this paper, we propose a deep learning based model and well-organized dataset for context-aware paper citation recommendation. Our model comprises a document encoder and a context encoder, which uses Graph Convolutional Networks (GCN) layer and Bidirectional Encoder Representations from Transformers (BERT), which is a pre-trained model of textual data. By modifying the related PeerRead dataset, we propose a new dataset called FullTextPeerRead containing context sentences to cited references and paper metadata. To the best of our knowledge, This dataset is the first well-organized dataset for context-aware paper recommendation. The results indicate that the proposed model with the proposed datasets can attain state-of-the-art performance and achieve a more than 28% improvement in mean average precision (MAP) and recall@k.
http://arxiv.org/abs/1903.06464
Affective Computing is a rapidly growing field spurred by advancements in artificial intelligence, but often, held back by the inability to translate psychological theories of emotion into tractable computational models. To address this, we propose a probabilistic programming approach to affective computing, which models psychological-grounded theories as generative models of emotion, and implements them as stochastic, executable computer programs. We first review probabilistic approaches that integrate reasoning about emotions with reasoning about other latent mental states (e.g., beliefs, desires) in context. Recently-developed probabilistic programming languages offer several key desidarata over previous approaches, such as: (i) flexibility in representing emotions and emotional processes; (ii) modularity and compositionality; (iii) integration with deep learning libraries that facilitate efficient inference and learning from large, naturalistic data; and (iv) ease of adoption. Furthermore, using a probabilistic programming framework allows a standardized platform for theory-building and experimentation: Competing theories (e.g., of appraisal or other emotional processes) can be easily compared via modular substitution of code followed by model comparison. To jumpstart adoption, we illustrate our points with executable code that researchers can easily modify for their own models. We end with a discussion of applications and future directions of the probabilistic programming approach.
http://arxiv.org/abs/1903.06445
A unified mathematical model for synchronisation and swarming has been proposed recently. Each system entity, called “swarmalator”, coordinates its internal phase and location with the other entities in a way that these two attributes are mutually coupled. This paper realises and studies, for the first time, the concept of swarmalators in technical systems. We adapt and extend the original model for its use on mobile robots and implement it in the Robot Operating System 2 (ROS 2). Simulations and experiments with small robots demonstrate the feasibility of the model and show its potential to be applied in real-world systems. All types of space-time patterns achieved in theory can be reproduced in practice. Applications can be found in monitoring, exploration, entertainment and art, among other domains.
http://arxiv.org/abs/1903.06440
The ability to quickly recognize and learn new visual concepts from limited samples enables humans to swiftly adapt to new environments. This ability is enabled by semantic associations of novel concepts with those that have already been learned and stored in memory. Computers can start to ascertain similar abilities by utilizing a semantic concept space. A concept space is a high-dimensional semantic space in which similar abstract concepts appear close and dissimilar ones far apart. In this paper, we propose a novel approach to one-shot learning that builds on this idea. Our approach learns to map a novel sample instance to a concept, relates that concept to the existing ones in the concept space and generates new instances, by interpolating among the concepts, to help learning. Instead of synthesizing new image instance, we propose to directly synthesize instance features by leveraging semantics using a novel auto-encoder network we call dual TriNet. The encoder part of the TriNet learns to map multi-layer visual features of deep CNNs, that is, multi-level concepts, to a semantic vector. In semantic space, we search for related concepts, which are then projected back into the image feature spaces by the decoder portion of the TriNet. Two strategies in the semantic space are explored. Notably, this seemingly simple strategy results in complex augmented feature distributions in the image feature space, leading to substantially better performance.
http://arxiv.org/abs/1804.05298
As Artificial Intelligence (AI) becomes an integral part of our life, the development of explainable AI, embodied in the decision-making process of an AI or robotic agent, becomes imperative. For a robotic teammate, the ability to generate explanations to explain its behavior is one of the key requirements of an explainable agency. Prior work on explanation generation focuses on supporting the reasoning behind the robot’s behavior. These approaches, however, fail to consider the cognitive effort needed to understand the received explanation. In particular, the human teammate is expected to understand any explanation provided before the task execution, no matter how much information is presented in the explanation. In this work, we argue that an explanation, especially complex ones, should be made in an online fashion during the execution, which helps to spread out the information to be explained and thus reducing the cognitive load of humans. However, a challenge here is that the different parts of an explanation are dependent on each other, which must be taken into account when generating online explanations. To this end, a general formulation of online explanation generation is presented. We base our explanation generation method in a model reconciliation setting introduced in our prior work. Our approach is evaluated both with human subjects in a standard planning competition (IPC) domain, using NASA Task Load Index (TLX), as well as in simulation with four different problems.
http://arxiv.org/abs/1903.06418
This paper describes technology developed to automatically grade Italian students (ages 9-16) on their English and German spoken language proficiency. The students’ spoken answers are first transcribed by an automatic speech recognition (ASR) system and then scored using a feedforward neural network (NN) that processes features extracted from the automatic transcriptions. In-domain acoustic models, employing deep neural networks (DNNs), are derived by adapting the parameters of an original out of domain DNN.
http://arxiv.org/abs/1903.06409
Interpretable classification models are built with the purpose of providing a comprehensible description of the decision logic to an external oversight agent. When considered in isolation, a decision tree, a set of classification rules, or a linear model, are widely recognized as human-interpretable. However, such models are generated as part of a larger analytical process. Bias in data collection and preparation, or in model’s construction may severely affect the accountability of the design process. We conduct an experimental study of the stability of interpretable models with respect to feature selection, instance selection, and model selection. Our conclusions should raise awareness and attention of the scientific community on the need of a stability impact assessment of interpretable models.
http://arxiv.org/abs/1810.09352
In autonomous driving community, numerous benchmarks have been established to assist the tasks of 3D/2D object detection, stereo vision, semantic/instance segmentation. However, the more meaningful dynamic evolution of the surrounding objects of ego-vehicle is rarely exploited, and lacks a large-scale dataset platform. To address this, we introduce BLVD, a large-scale 5D semantics benchmark which does not concentrate on the static detection or semantic/instance segmentation tasks tackled adequately before. Instead, BLVD aims to provide a platform for the tasks of dynamic 4D (3D+temporal) tracking, 5D (4D+interactive) interactive event recognition and intention prediction. This benchmark will boost the deeper understanding of traffic scenes than ever before. We totally yield 249,129 3D annotations, 4,902 independent individuals for tracking with the length of overall 214,922 points, 6,004 valid fragments for 5D interactive event recognition, and 4,900 individuals for 5D intention prediction. These tasks are contained in four kinds of scenarios depending on the object density (low and high) and light conditions (daytime and nighttime). The benchmark can be downloaded from our project site https://github.com/VCCIV/BLVD/.
http://arxiv.org/abs/1903.06405