检索结果-内蒙古大学图书馆

Render in-between: Motion guided video synthesis for action interpolation

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Ho, Hsuan-I Chen, Xu Song, Jie Hilliges, Otmar Department of Computer Science ETH Zürich Max Planck Institute for Intelligent Systems Tübingen Germany

Frame rate greatly impacts the style and viewing experience of a video. Especially for footage that depicts fast motion, higher frame rate equates to smoother imagery and reduced motion blur artifacts. Upsampling of low-frame-rate videos of human activity is an interesting yet challenging task with many potential applications ranging from gaming to entertainment and sports broadcasting. The main difficulty in synthesizing video frames in this setting stems from the highly complex and non-linear nature of human motion and the complex appearance and texture of the body. We propose to address these issues in a motion-guided frame-upsampling framework that is capable of producing realistic human motion and appearance. A novel motion model is trained to interpolate the non-linear skeletal motion between frames by leveraging a large-scale motion-capture dataset (AMASS). The high-frame-rate pose predictions are then used by a neural rendering pipeline to produce the full-frame output, taking the pose and background consistency into consideration. Our pipeline only requires low-frame-rate videos and unpaired human motion data but does not require high-frame-rate videos for training. Furthermore, we contribute the first evaluation dataset that consists of high-quality and high-frame-rate videos of human activities for this task. Compared with state-of-the-art video interpolation techniques, our method produces interpolated frames with better quality and accuracy, which is evident by state-of-the-art results on pixel-level, distributional metrics and comparative user evaluations. Our code and the collected dataset are available at https://***/Render-In-Between. Copyright © 2021, The Authors. All rights reserved.

关键词： Signal sampling

Planning from pixels in environments with combinatorially hard search spaces 21

学校读者我要写书评

暂无评论

Planning from pixels in environments with combinatorially ha...

Proceedings of the 35th International Conference on Neural Information Processing systems

作者： Marco Bagatella Mirek Olšák Michal Rolínek Georg Martius Max Planck Institute for Intelligent Systems Tübingen Germany Computer Science Department University Innsbruck Austria

ISBN: (纸本)9781713845393

The ability to form complex plans based on raw visual input is a litmus test for current capabilities of artificial intelligence, as it requires a seamless combination of visual processing and abstract algorithmic execution, two traditionally separate areas of computer science. A recent surge of interest in this field brought advances that yield good performance in tasks ranging from arcade games to continuous control; these methods however do not come without significant issues, such as limited generalization capabilities and difficulties when dealing with combina-torially hard planning instances. Our contribution is two-fold: (i) we present a method that learns to represent its environment as a latent graph and leverages state reidentification to reduce the complexity of finding a good policy from exponential to linear (ii) we introduce a set of lightweight environments with an underlying discrete combinatorial structure in which planning is challenging even for humans. Moreover, we show that our methods achieves strong empirical generalization to variations in the environment, even across highly disadvantaged regimes, such as "one-shot" planning, or in an offline RL paradigm which only provides low-quality trajectories.

关键词：

Sampling Strategies for Exploratory Landscape Analysis of Bi-Objective Problems

学校读者我要写书评

暂无评论

Sampling Strategies for Exploratory Landscape Analysis of Bi...

International Conference on Computational science and Computational Intelligence (CSCI)

作者： Pavel Krömer Vojtěch Uher Andrejaana Andova Tea Tušar Bogdan Filipič Department of Computer Science VSB - Technical University of Ostrava Ostrava-Poruba Czech Republic Department of Intelligent Systems Jožef Stefan Institute and Jožef Stefan International Postgraduate School Ljubljana Slovenia

Exploratory landscape analysis (ELA) is a popular method for the understanding of complex, often black-box optimization problems. It tries to approximate and describe the surfaces formed by the fitness and other characteristic values associated with problem solutions on top of the multi-dimensional solution spaces. Sampling is the initial step of the ELA pipeline. It is a strategy for selecting a limited number of solutions, i.e., points in the multi-dimensional solution space, for which the fitness function(s) are evaluated. Consequently, the fitness landscape is approximated and its properties are drawn from these fitness values. In this work, the properties and the impact of various sampling strategies on the analysis of the fitness landscape are studied in the context of bi-objective optimization. Extensive computational experiments show that the use of different sampling strategies affects both the value of high-level landscape features and their usability for problem classification. The results also demonstrate that the magnitude and significance of the impact depend on problem dimension and sample size.

关键词：

Constrained Reinforcement Learning for Dynamic Material Handling

学校读者我要写书评

暂无评论

Constrained Reinforcement Learning for Dynamic Material Hand...

International Joint Conference on Neural Networks (IJCNN)

作者： Chengpeng Hu Ziming Wang Jialin Liu Junyi Wen Bifei Mao Xin Yao Research Institute of Trustworthy Autonomous Systems (RITAS) Southern University of Science and Technology Shenzhen China Department of Computer Science and Engineering Guangdong Key Laboratory of Brain-inspired Intelligent Computation Southern University of Science and Technology Shenzhen China Trustworthiness Theory Research Center Huawei Technologies Co. Ltd Shenzhen China

As one of the core parts of flexible manufacturing systems, material handling involves storage and transportation of materials between workstations with automated vehicles. The improvement in material handling can impulse the overall efficiency of the manufacturing system. However, the occurrence of dynamic events during the optimisation of task arrangements poses a challenge that requires adaptability and effectiveness. In this paper, we aim at the scheduling of automated guided vehicles for dynamic material handling. Motivated by some real-world scenarios, unknown new tasks and unexpected vehicle breakdowns are regarded as dynamic events in our problem. We formulate the problem as a constrained Markov decision process which takes into account tardiness and available vehicles as cumulative and instantaneous constraints, respectively. An adaptive constrained reinforcement learning algorithm that combines Lagrangian relaxation and invalid action masking, named RCPOM, is proposed to address the problem with two hybrid constraints. Moreover, a gym-like dynamic material handling simulator, named DMH-GYM, is developed and equipped with diverse problem instances, which can be used as benchmarks for dynamic material handling. Experimental results on the problem instances demonstrate the outstanding performance of our proposed approach compared with eight state-of-the-art constrained and non-constrained reinforcement learning algorithms, and widely used dispatching rules for material handling.

关键词：

Self-supervised Reinforcement Learning with Independently Controllable Subgoals

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Zadaianchuk, Andrii Martius, Georg Yang, Fanny Max Planck Institute for Intelligent Systems Tübingen Germany Department of Computer Science ETH Zurich Switzerland

To successfully tackle challenging manipulation tasks, autonomous agents must learn a diverse set of skills and how to combine them. Recently, self-supervised agents that set their own abstract goals by exploiting the discovered structure in the environment were shown to perform well on many different tasks. In particular, some of them were applied to learn basic manipulation skills in compositional multi-object environments. However, these methods learn skills without taking the dependencies between objects into account. Thus, the learned skills are difficult to combine in realistic environments. We propose a novel self-supervised agent that estimates relations between environment components and uses them to independently control different parts of the environment state. In addition, the estimated relations between objects can be used to decompose a complex goal into a compatible sequence of subgoals. We show that, by using this framework, an agent can efficiently and automatically learn manipulation tasks in multi-object environments with different relations between objects. © 2021, CC BY.

关键词： Autonomous agents

Automatic Threat Detection in Baggage Security Imagery using Deep Learning Models 15

学校读者我要写书评

暂无评论

Automatic Threat Detection in Baggage Security Imagery using...

15th IEEE International Conference on Industrial and Information systems, ICIIS 2020

作者： Mithal, Aditya Baser, Manit Dhiraj Birla Institute of Technology and Science Department of Computer Science Engineering Pilani India CSIR-Central Electronics Engineering Research Institute Intelligent Systems Group Pilani India

ISBN: (纸本)9781728185248

Automating object detection for surveillance purpose and threat detection is beneficial as it may compensate for the human error and will save time, which is of significant economic value. For the end-to-end classification process and feature extraction, the CNN approach requires large amounts of data. To overcome this limited availability of data, we have presented a transfer learning approach with various object detection models for single and multiple detections on two types of the dataset: Single-channelled (GDXray dataset) and Multichanneled(SIXray dataset). We have presented comparisons between the various models(Faster R-CNN with ResNet50, SSD with VGG16, YOLOv3 with ResNet50, and RetinaNet with ResNet50). The best results were achieved on Faster-RCNN(ResNet50) with 0.966 mAP for the four-class object detection problem(GDXray Dataset) and 0.845 mAP for the two-class object detection problem(SIXray Dataset). © 2020 IEEE.

关键词： Object detection

Planning from Pixels in Environments with Combinatorially Hard Search Spaces

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Bagatella, Marco Olšák, Mirek Rolínek, Michal Martius, Georg Max Planck Institute for Intelligent Systems Tübingen Germany Computer Science Department University Innsbruck Austria

The ability to form complex plans based on raw visual input is a litmus test for current capabilities of artificial intelligence, as it requires a seamless combination of visual processing and abstract algorithmic execution, two traditionally separate areas of computer science. A recent surge of interest in this field brought advances that yield good performance in tasks ranging from arcade games to continuous control;these methods however do not come without significant issues, such as limited generalization capabilities and difficulties when dealing with combinatorially hard planning instances. Our contribution is two-fold: (i) we present a method that learns to represent its environment as a latent graph and leverages state reidentification to reduce the complexity of finding a good policy from exponential to linear (ii) we introduce a set of lightweight environments with an underlying discrete combinatorial structure in which planning is challenging even for humans. Moreover, we show that our methods achieves strong empirical generalization to variations in the environment, even across highly disadvantaged regimes, such as "one-shot" planning, or in an offline RL paradigm which only provides low-quality trajectories. Copyright © 2021, The Authors. All rights reserved.

关键词： Machine learning

AircraftVerse: A Large-Scale Multimodal Dataset of Aerial Vehicle Designs

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Cobb, Adam D. Roy, Anirban Elenius, Daniel Heim, F. Michael Swenson, Brian Whittington, Sydney Walker, James D. Bapty, Theodore Hite, Joseph Ramani, Karthik McComb, Christopher Jha, Susmit Computer Science Laboratory SRI International United States Mechanical Engineering Division Southwest Research Institute United States Intelligent Systems Division Southwest Research Institute United States Institute of Software Integrated Systems Vanderbilt University United States School of Mechanical Engineering Purdue University United States Department of Mechanical Engineering Carnegie Mellon University United States

We present AircraftVerse, a publicly available aerial vehicle design dataset. Aircraft design encompasses different physics domains and, hence, multiple modalities of representation. The evaluation of these cyber-physical system (CPS) designs requires the use of scientific analytical and simulation models ranging from computer-aided design tools for structural and manufacturing analysis, computational fluid dynamics tools for drag and lift computation, battery models for energy estimation, and simulation models for flight control and dynamics. AircraftVerse contains 27,714 diverse air vehicle designs - the largest corpus of engineering designs with this level of complexity. Each design comprises the following artifacts: a symbolic design tree describing topology, propulsion subsystem, battery subsystem, and other design details;a STandard for the Exchange of Product (STEP) model data;a 3D CAD design using a stereolithography (STL) file format;a 3D point cloud for the shape of the design;and evaluation results from high fidelity state-of-the-art physics models that characterize performance metrics such as maximum flight distance and hover-time. We also present baseline surrogate models that use different modalities of design representation to predict design performance metrics, which we provide as part of our dataset release. Finally, we discuss the potential impact of this dataset on the use of learning in aircraft design and, more generally, in CPS. AircraftVerse is accompanied by a data card, and it is released under Creative Commons Attribution-ShareAlike (CC BY-SA) license. The dataset is hosted at https://***/record/6525446, baseline models and code at https://***/SRI-CSL/AircraftVerse, and the dataset description at https://***/. © 2023, CC BY-SA.

关键词： computer aided design

ANALYZING THE PERFORMANCE OF DEEP ENCODER-DECODER NETWORKS AS SURROGATES FOR A DIFFUSION EQUATION

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Toledo-Marín, J. Quetzalcóatl Glazier, James A. Fox, Geoffrey University of British Columbia BC Children’s Hospital Research Institute VancouverBC Canada Biocomplexity Institute Department of Intelligent Systems Engineering Indiana University BloomingtonIN47408 United States University of Virginia Computer Science and Biocomplexity Institute 994 Research Park Blvd CharlottesvilleVA22911 United States

Neural networks (NNs) have been demonstrated to be a viable alternative to traditional direct numerical evaluation algorithms, with the potential to accelerate computational time by several orders of magnitude. In the present paper we study the use of encoder-decoder convolutional neural network (CNN) algorithms as surrogates for steady-state diffusion solvers. The construction of such surrogates requires the selection of an appropriate task, network architecture, training set structure and size, loss function, and training algorithm hyperparameters. It is well known that each of these factors can have a significant impact on the performance of the resultant model. Our approach employs an encoder-decoder CNN architecture, which we posit is particularly well-suited for this task due to its ability to effectively transform data, as opposed to merely compressing it. We systematically evaluate a range of loss functions, hyperparameters, and training set sizes. Our results indicate that increasing the size of the training set has a substantial effect on reducing performance fluctuations and overall error. Additionally, we observe that the performance of the model exhibits a logarithmic dependence on the training set size. Furthermore, we investigate the effect on model performance by using different subsets of data with varying features. Our results highlight the importance of sampling the configurational space in an optimal manner, as this can have a significant impact on the performance of the model and the required training time. In conclusion, our results suggest that training a model with a pre-determined error performance bound is not a viable approach, as it does not guarantee that edge cases with errors larger than the bound do not exist. Furthermore, as most surrogate tasks involve a high dimensional landscape, an ever increasing training set size is, in principle, needed, however it is not a practical solution. © 2023, CC BY.

关键词： Errors