This paper presents a generalized framework for the simulation of multiple robots and drones in highly realistic models of natural environments. The proposed simulation architecture uses the Unreal Engine4 for generating both optical and depth sensor outputs from any position and orientation within the environment and provides several key domain specific simulation capabilities. Various components and functionalities of the system have been discussed in detail. The simulation engine also allows users to test and validate a wide range of computer vision algorithms involving different drone configurations under many types of environmental effects such as wind gusts. The paper demonstrates the effectiveness of the system by giving experimental results for a test scenario where one drone tracks the simulated motion of another in a complex natural environment.
The region-based Convolutional Neural Network (CNN) detectors such as Faster R-CNN or R-FCN have already shown promising results for object detection by combining the region proposal subnetwork and the classification subnetwork together. Although R-FCN has achieved higher detection speed while keeping the detection performance, the global structure information is ignored by the position-sensitive score maps. To fully explore the local and global properties, in this paper, we propose a novel fully convolutional network, named as CoupleNet, to couple the global structure with local parts for object detection. Specifically, the object proposals obtained by the Region Proposal Network (RPN) are fed into the the coupling module which consists of two branches. One branch adopts the position-sensitive RoI (PSRoI) pooling to capture the local part information of the object, while the other employs the RoI pooling to encode the global and context information. Next, we design different coupling strategies and normalization ways to make full use of the complementary advantages between the global and local branches. Extensive experiments demonstrate the effectiveness of our approach. We achieve state-of-the-art results on all three challenging datasets, i.e. a mAP of 82.7% on VOC07, 80.4% on VOC12, and 34.4% on COCO. Codes will be made publicly available.
This paper presents a framework for localization or grounding of phrases in images using a large collection of linguistic and visual cues. We model the appearance, size, and position of entity bounding boxes, adjectives that contain attribute information, and spatial relationships between pairs of entities connected by verbs or prepositions. Special attention is given to relationships between people and clothing or body part mentions, as they are useful for distinguishing individuals. We automatically learn weights for combining these cues and at test time, perform joint inference over all phrases in a caption. The resulting system produces state of the art performance on phrase localization on the Flickr30k Entities dataset and visual relationship detection on the Stanford VRD dataset.
In this contribution, a software system for computer-aided position planning of miniplates to treat facial bone defects is proposed. The intra-operatively used bone plates have to be passively adapted on the underlying bone contours for adequate bone fragment stabilization. However, this procedure can lead to frequent intra-operatively performed material readjustments especially in complex surgical cases. Our approach is able to fit a selection of common implant models on the surgeon's desired position in a 3D computer model. This happens with respect to the surrounding anatomical structures, always including the possibility of adjusting both the direction and the position of the used osteosynthesis material. By using the proposed software, surgeons are able to pre-plan the out coming implant in its form and morphology with the aid of a computer-visualized model within a few minutes. Further, the resulting model can be stored in STL file format, the commonly used format for 3D printing. Using this technology, surgeons are able to print the virtual generated implant, or create an individually designed bending tool. This method leads to adapted osteosynthesis materials according to the surrounding anatomy and requires further a minimum amount of money and time.
We propose here an extended attention model for sequence-to-sequence recurrent neural networks (RNNs) designed to capture (pseudo-)periods in time series. This extended attention model can be deployed on top of any RNN and is shown to yield state-of-the-art performance for time series forecasting on several univariate and multivariate time series.
We propose several cost functions for registration of shapes encoded with Euclidean and/or non-Euclidean information (unit vectors). Our framework is assessed for estimation of both rigid and non-rigid transformations between the target and model shapes corresponding to 2D contours and 3D surfaces. The experimental results obtained confirm that using the combination of a point's position and unit normal vector in a cost function can enhance the registration results compared to state of the art methods.
This paper presents a comprehensive review on regression-based method for human pose estimation. The problem of human pose estimation has been intensively studied and enabled many application from entertainment to training. Traditional methods often rely on color image only which cannot completely ambiguity of joint 3D position, especially in the complex context. With the popularity of depth sensors, the precision of 3D estimation has significant improvement. In this paper, we give a detailed analysis of state-of-the-art on human pose estimation, including depth image based and RGB-D based approaches. The experimental results demonstrate their advantages and limitation for different scenarios.
One key challenge in talent search is to translate complex criteria of a hiring position into a search query, while it is relatively easy for a searcher to list examples of suitable candidates for a given position. To improve search efficiency, we propose the next generation of talent search at LinkedIn, also referred to as Search By Ideal Candidates. In this system, a searcher provides one or several ideal candidates as the input to hire for a given position. The system then generates a query based on the ideal candidates and uses it to retrieve and rank results. Shifting from the traditional Query-By-Keyword to this new Query-By-Example system poses a number of challenges: How to generate a query that best describes the candidates? When moving to a completely different paradigm, how does one leverage previous product logs to learn ranking models and/or evaluate the new system with no existing usage logs? Finally, given the different nature between the two search paradigms, the ranking features typically used for Query-By-Keyword systems might not be optimal for Query-By-Example. This paper describes our approach to solving these challenges. We present experimental results confirming the effectiveness of the proposed solution, particularly on query building and search ranking tasks. As of writing this paper, the new system has been available to all LinkedIn members.
Recent robotic manipulation competitions have highlighted that sophisticated robots still struggle to achieve fast and reliable perception of task-relevant objects in complex, realistic scenarios. To improve these systems' perceptive speed and robustness, we present SegICP, a novel integrated solution to object recognition and pose estimation. SegICP couples convolutional neural networks and multi-hypothesis point cloud registration to achieve both robust pixel-wise semantic segmentation as well as accurate and real-time 6-DOF pose estimation for relevant objects. Our architecture achieves 1cm position error and <5^\circ$ angle error in real time without an initial seed. We evaluate and benchmark SegICP against an annotated dataset generated by motion capture.
We present software to detect noise in eye position signals from video-based eye-tracking systems that depend on accurate pupil and corneal reflection position estimation. When such systems transiently fail to properly detect the pupil or the corneal reflection due to occlusion from eyelids, eye lashes or various shadows, the estimated gaze position is false. This produces an artifactual signal in the position trace that is rapidly, irregularly oscillating between true and false gaze positions. We refer to this noise as RIONEPS (Rapid Irregularly Oscillating Noise of the Eye Position Signal). Our method for detecting these periods automatically is based on an estimate of the relative inefficiency of the eye position signal. We look for RIONEPS in the horizontal and vertical traces separately, and although we typically use it offline, it is suitable to adaptation for real time use. This method requires a threshold to be set, and although we provide some guidance, thresholds will have to be estimated empirically.
The auto-encoder method is a type of dimensionality reduction method. A mapping from a vector to a descriptor that represents essential information can be automatically generated from a set of vectors without any supervising information. However, an image and its spatially shifted version are encoded into different descriptors by an existing ordinary auto-encoder because each descriptor includes a spatial subpattern and its position. To generate a descriptor representing a spatial subpattern in an image, we need to normalize its spatial position in the images prior to training an ordinary auto-encoder; however, such a normalization is generally difficult for images without obvious standard positions. We propose a transform invariant auto-encoder and an inference model of transform parameters. By the proposed method, we can separate an input into a transform invariant descriptor and transform parameters. The proposed method can be applied to various auto-encoders without requiring any special modules or labeled training samples. By applying it to shift transforms, we can achieve a shift invariant auto-encoder that can extract a typical spatial subpattern independent of its relative position in a window. In addition, we can achieve a model that can infer shift parameters required to restore the input from the typical subpattern. As an example of the proposed method, we demonstrate that a descriptor generated by a shift invariant auto-encoder can represent a typical spatial subpattern. In addition, we demonstrate the imitation of a human hand by a robot hand as an example of a regression based on spatial subpatterns.
Optical see-through head-mounted displays (OST HMDs) are a major output medium for Augmented Reality, which have seen significant growth in popularity and usage among the general public due to the growing release of consumer-oriented models, such as the Microsoft Hololens. Unlike Virtual Reality headsets, OST HMDs inherently support the addition of computer-generated graphics directly into the light path between a user's eyes and their view of the physical world. As with most Augmented and Virtual Reality systems, the physical position of an OST HMD is typically determined by an external or embedded 6-Degree-of-Freedom tracking system. However, in order to properly render virtual objects, which are perceived as spatially aligned with the physical environment, it is also necessary to accurately measure the position of the user's eyes within the tracking system's coordinate frame. For over 20 years, researchers have proposed various calibration methods to determine this needed eye position. However, to date, there has not been a comprehensive overview of these procedures and their requirements. Hence, this paper surveys the field of calibration methods for OST HMDs. Specifically, it provides insights into the fundamentals of calibration techniques, and presents an overview of both manual and automatic approaches, as well as evaluation methods and metrics. Finally, it also identifies opportunities for future research. % relative to the tracking coordinate system, and, hence, its position in 3D space.
Our understanding of the world depends highly on our capacity to produce intuitive and simplified representations which can be easily used to solve problems. We reproduce this simplification process using a neural network to build a low dimensional state representation of the world from images acquired by a robot. As in Jonschkowski et al. 2015, we learn in an unsupervised way using prior knowledge about the world as loss functions called robotic priors and extend this approach to high dimension richer images to learn a 3D representation of the hand position of a robot from RGB images. We propose a quantitative evaluation of the learned representation using nearest neighbors in the state space that allows to assess its quality and show both the potential and limitations of robotic priors in realistic environments. We augment image size, add distractors and domain randomization, all crucial components to achieve transfer learning to real robots. Finally, we also contribute a new prior to improve the robustness of the representation. The applications of such low dimensional state representation range from easing reinforcement learning (RL) and knowledge transfer across tasks, to facilitating learning from raw data with more efficient and compact high level representations. The results show that the robotic prior approach is able to extract high level representation as the 3D position of an arm and organize it into a compact and coherent space of states in a challenging dataset.
We present a direct method to calculate a 6DoF pose change of a monocular camera for mobile navigation. The calculated pose is estimated up to a constant unknown scale parameter that is kept constant over the entire reconstruction process. This method allows a direct cal- culation of the metric position and rotation without any necessity to fuse the information in a probabilistic approach over longer frame sequence as it is the case in most currently used VSLAM approaches. The algorithm provides two novel aspects to the field of monocular navigation. It allows a direct pose estimation without any a-priori knowledge about the world directly from any two images and it provides a quality measure for the estimated motion parameters that allows to fuse the resulting information in Kalman Filters. We present the mathematical formulation of the approach together with experimental validation on real scene images.
Face detection has achieved great success using the region-based methods. In this report, we propose a region-based face detector applying deep networks in a fully convolutional fashion, named Face R-FCN. Based on Region-based Fully Convolutional Networks (R-FCN), our face detector is more accurate and computational efficient compared with the previous R-CNN based face detectors. In our approach, we adopt the fully convolutional Residual Network (ResNet) as the backbone network. Particularly, We exploit several new techniques including position-sensitive average pooling, multi-scale training and testing and on-line hard example mining strategy to improve the detection accuracy. Over two most popular and challenging face detection benchmarks, FDDB and WIDER FACE, Face R-FCN achieves superior performance over state-of-the-arts.
Robotic learning in simulation environments provides a faster, more scalable, and safer training methodology than learning directly with physical robots. Also, synthesizing images in a simulation environment for collecting large-scale image data is easy, whereas capturing camera images in the real world is time consuming and expensive. However, learning from only synthetic images may not achieve the desired performance in real environments due to the gap between synthetic and real images. We thus propose a method that transfers learned capability of detecting object position from a simulation environment to the real world. Our method enables us to use only a very limited dataset of real images while leveraging a large dataset of synthetic images using multiple variational autoencoders. It detects object positions 6 to 7 times more precisely than the baseline of directly learning from the dataset of the real images. Object position estimation under varying environmental conditions forms one of the underlying requirement for standard robotic manipulation tasks. We show that the proposed method performs robustly in different lighting conditions or with other distractor objects present for this requirement. Using this detected object position, we transfer pick-and-place or reaching tasks learned in a simulation environment to an actual physical robot without re-training.
Light scattered from multiple surfaces can be used to retrieve information of hidden environments. However, full three-dimensional retrieval of an object hidden from view by a wall has only been achieved with scanning systems and requires intensive computational processing of the retrieved data. Here we use a non-scanning, single-photon single-pixel detector in combination with an artificial neural network: this allows us to locate the position and to also simultaneously provide the actual identity of a hidden person, chosen from a database of people (N=3). Artificial neural networks applied to specific computational imaging problems can therefore enable novel imaging capabilities with hugely simplified hardware and processing times
Spectral embedding is a procedure which can be used to obtain vector representations of the nodes of a graph. This paper proposes a generalisation of the latent position network model known as the random dot product graph, to allow interpretation of those vector representations as latent position estimates. The generalisation is needed to model heterophilic connectivity (e.g., `opposites attract') and to cope with negative eigenvalues more generally. We show that, whether the adjacency or normalised Laplacian matrix is used, spectral embedding produces uniformly consistent latent position estimates with asymptotically Gaussian error (up to identifiability). The standard and mixed membership stochastic block models are special cases in which the latent positions take only $K$ distinct vector values, representing communities, or live in the $(K-1)$-simplex with those vertices, respectively. Under the stochastic block model, our theory suggests spectral clustering using a Gaussian mixture model (rather than $K$-means) and, under mixed membership, fitting the minimum volume enclosing simplex, existing recommendations previously only supported under non-negative-definite assumptions. Empirical improvements in link prediction (over the random dot product graph), and the potential to uncover richer latent structure (than posited under the standard or mixed membership stochastic block models) are demonstrated in a cyber-security example.
High precision assembly of mechanical parts requires accuracy exceeding the robot precision. Conventional part mating methods used in the current manufacturing requires tedious tuning of numerous parameters before deployment. We show how the robot can successfully perform a tight clearance peg-in-hole task through training a recurrent neural network with reinforcement learning. In addition to saving the manual effort, the proposed technique also shows robustness against position and angle errors for the peg-in-hole task. The neural network learns to take the optimal action by observing the robot sensors to estimate the system state. The advantages of our proposed method is validated experimentally on a 7-axis articulated robot arm.
In this paper, we propose a statistical test to determine whether a given word is used as a polysemic word or not. The statistic of the word in this test roughly corresponds to the fluctuation in the senses of the neighboring words a nd the word itself. Even though the sense of a word corresponds to a single vector, we discuss how polysemy of the words affects the position of vectors. Finally, we also explain the method to detect this effect.