PyTorchKR Papers

최신 논문

총 5,575편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2006.13194

Instant 3D Object Tracking with Applications in Augmented Reality

Tracking object poses in 3D is a crucial building block for Augmented Reality applications. We propose an instant motion tracking system that tracks an object's pose in space (represented by its 3D bounding box) in real-time on mobile devices. Our system does not require any prior sensory calibration or initialization to function. We employ a deep neural network to detect objects and estimate their initial 3D pose. Then the estimated pose is tracked using a robust planar tracker. Our tracker is capable of performing relative-scale 9-DoF tracking in real-time on mobile devices. By combining use of CPU and GPU efficiently, we achieve 26-FPS+ performance on mobile devices.

arXiv:2006.12791

Increased-Range Unsupervised Monocular Depth Estimation

Unsupervised deep learning methods have shown promising performance for single-image depth estimation. Since most of these methods use binocular stereo pairs for self-supervision, the depth range is generally limited. Small-baseline stereo pairs provide small depth range but handle occlusions well. On the other hand, stereo images acquired with a wide-baseline rig cause occlusions-related errors in the near range but estimate depth well in the far range. In this work, we propose to integrate the advantages of the small and wide baselines. By training the network using three horizontally aligned views, we obtain accurate depth predictions for both close and far ranges. Our strategy allows to infer multi-baseline depth from a single image. This is unlike previous multi-baseline systems which employ more than two cameras. The qualitative and quantitative results show the superior performance of multi-baseline approach over previous stereo-based monocular methods. For 0.1 to 80 meters depth range, our approach decreases the absolute relative error of depth by 24% compared to Monodepth2. Our approach provides 21 frames per second on a single Nvidia1080 GPU, making it useful for practical applications.

arXiv:2006.12671

AFDet: Anchor Free One Stage 3D Object Detection

High-efficiency point cloud 3D object detection operated on embedded systems is important for many robotics applications including autonomous driving. Most previous works try to solve it using anchor-based detection methods which come with two drawbacks: post-processing is relatively complex and computationally expensive; tuning anchor parameters is tricky. We are the first to address these drawbacks with an anchor free and Non-Maximum Suppression free one stage detector called AFDet. The entire AFDet can be processed efficiently on a CNN accelerator or a GPU with the simplified post-processing. Without bells and whistles, our proposed AFDet performs competitively with other one stage anchor-based methods on KITTI validation set and Waymo Open Dataset validation set.

arXiv:2006.11972

Hippo: Taming Hyper-parameter Optimization of Deep Learning with Stage Trees

Hyper-parameter optimization is crucial for pushing the accuracy of a deep learning model to its limits. A hyper-parameter optimization job, referred to as a study, involves numerous trials of training a model using different training knobs, and therefore is very computation-heavy, typically taking hours and days to finish. We observe that trials issued from hyper-parameter optimization algorithms often share common hyper-parameter sequence prefixes. Based on this observation, we propose Hippo, a hyper-parameter optimization system that removes redundancy in the training process to reduce the overall amount of computation significantly. Instead of executing each trial independently as in existing hyper-parameter optimization systems, Hippo breaks down the hyper-parameter sequences into stages and merges common stages to form a tree of stages (called a stage-tree), then executes a stage once per tree on a distributed GPU server environment. Hippo is applicable to not only single studies, but multi-study scenarios as well, where multiple studies of the same model and search space can be formulated as trees of stages. Evaluations show that Hippo's stage-based execution strategy outperforms trial-based methods such as Ray Tune for several models and hyper-parameter optimization algorithms, reducing GPU-hours and end-to-end training time significantly.

arXiv:2006.11751공식 코드 · Papers with Code

Sample Factory: Egocentric 3D Control from Pixels at 100000 FPS with Asynchronous Reinforcement Learning

Increasing the scale of reinforcement learning experiments has allowed researchers to achieve unprecedented results in both training sophisticated agents for video games, and in sim-to-real transfer for robotics. Typically such experiments rely on large distributed systems and require expensive hardware setups, limiting wider access to this exciting area of research. In this work we aim to solve this problem by optimizing the efficiency and resource utilization of reinforcement learning algorithms instead of relying on distributed computation. We present the "Sample Factory", a high-throughput training system optimized for a single-machine setting. Our architecture combines a highly efficient, asynchronous, GPU-based sampler with off-policy correction techniques, allowing us to achieve throughput higher than $10^5$ environment frames/second on non-trivial control problems in 3D without sacrificing sample efficiency. We extend Sample Factory to support self-play and population-based training and apply these techniques to train highly capable agents for a multiplayer first-person shooter game. The source code is available at https://github.com/alex-petrenko/sample-factory

arXiv:2006.11718

Pose Trainer: Correcting Exercise Posture using Pose Estimation

Fitness exercises are very beneficial to personal health and fitness; however, they can also be ineffective and potentially dangerous if performed incorrectly by the user. Exercise mistakes are made when the user does not use the proper form, or pose. In our work, we introduce Pose Trainer, an application that detects the user's exercise pose and provides personalized, detailed recommendations on how the user can improve their form. Pose Trainer uses the state of the art in pose estimation to detect a user's pose, then evaluates the vector geometry of the pose through an exercise to provide useful feedback. We record a dataset of over 100 exercise videos of correct and incorrect form, based on personal training guidelines, and build geometric-heuristic and machine learning algorithms for evaluation. Pose Trainer works on four common exercises and supports any Windows or Linux computer with a GPU.

arXiv:2006.12085공식 코드 · Papers with Code

Split to Be Slim: An Overlooked Redundancy in Vanilla Convolution

Many effective solutions have been proposed to reduce the redundancy of models for inference acceleration. Nevertheless, common approaches mostly focus on eliminating less important filters or constructing efficient operations, while ignoring the pattern redundancy in feature maps. We reveal that many feature maps within a layer share similar but not identical patterns. However, it is difficult to identify if features with similar patterns are redundant or contain essential details. Therefore, instead of directly removing uncertain redundant features, we propose a \textbf{sp}lit based \textbf{conv}olutional operation, namely SPConv, to tolerate features with similar patterns but require less computation. Specifically, we split input feature maps into the representative part and the uncertain redundant part, where intrinsic information is extracted from the representative part through relatively heavy computation while tiny hidden details in the uncertain redundant part are processed with some light-weight operation. To recalibrate and fuse these two groups of processed features, we propose a parameters-free feature fusion module. Moreover, our SPConv is formulated to replace the vanilla convolution in a plug-and-play way. Without any bells and whistles, experimental results on benchmarks demonstrate SPConv-equipped networks consistently outperform state-of-the-art baselines in both accuracy and inference time on GPU, with FLOPs and parameters dropped sharply.

arXiv:2006.10901공식 코드 · Papers with Code

Sparse GPU Kernels for Deep Learning

Scientific workloads have traditionally exploited high levels of sparsity to accelerate computation and reduce memory requirements. While deep neural networks can be made sparse, achieving practical speedups on GPUs is difficult because these applications have relatively moderate levels of sparsity that are not sufficient for existing sparse kernels to outperform their dense counterparts. In this work, we study sparse matrices from deep learning applications and identify favorable properties that can be exploited to accelerate computation. Based on these insights, we develop high-performance GPU kernels for two sparse matrix operations widely applicable in neural networks: sparse matrix-dense matrix multiplication and sampled dense-dense matrix multiplication. Our kernels reach 27% of single-precision peak on Nvidia V100 GPUs. Using our kernels, we demonstrate sparse Transformer and MobileNet models that achieve 1.2-2.1x speedups and up to 12.8x memory savings without sacrificing accuracy.

arXiv:2006.10350공식 코드 · Papers with Code

Kernel methods through the roof: handling billions of points efficiently

Kernel methods provide an elegant and principled approach to nonparametric learning, but so far could hardly be used in large scale problems, since na\"ive implementations scale poorly with data size. Recent advances have shown the benefits of a number of algorithmic ideas, for example combining optimization, numerical linear algebra and random projections. Here, we push these efforts further to develop and test a solver that takes full advantage of GPU hardware. Towards this end, we designed a preconditioned gradient solver for kernel methods exploiting both GPU acceleration and parallelization with multiple GPUs, implementing out-of-core variants of common linear algebra operations to guarantee optimal hardware utilization. Further, we optimize the numerical precision of different operations and maximize efficiency of matrix-vector multiplications. As a result we can experimentally show dramatic speedups on datasets with billions of points, while still guaranteeing state of the art performance. Additionally, we make our software available as an easy to use library.

arXiv:2006.10201공식 코드 · Papers with Code

On the Role of Sparsity and DAG Constraints for Learning Linear DAGs

Learning graphical structures based on Directed Acyclic Graphs (DAGs) is a challenging problem, partly owing to the large search space of possible graphs. A recent line of work formulates the structure learning problem as a continuous constrained optimization task using the least squares objective and an algebraic characterization of DAGs. However, the formulation requires a hard DAG constraint and may lead to optimization difficulties. In this paper, we study the asymptotic role of the sparsity and DAG constraints for learning DAG models in the linear Gaussian and non-Gaussian cases, and investigate their usefulness in the finite sample regime. Based on the theoretical results, we formulate a likelihood-based score function, and show that one only has to apply soft sparsity and DAG constraints to learn a DAG equivalent to the ground truth DAG. This leads to an unconstrained optimization problem that is much easier to solve. Using gradient-based optimization and GPU acceleration, our procedure can easily handle thousands of nodes while retaining a high accuracy. Extensive experiments validate the effectiveness of our proposed method and show that the DAG-penalized likelihood objective is indeed favorable over the least squares one with the hard DAG constraint.

arXiv:2006.09628

Visor: Privacy-Preserving Video Analytics as a Cloud Service

Video-analytics-as-a-service is becoming an important offering for cloud providers. A key concern in such services is privacy of the videos being analyzed. While trusted execution environments (TEEs) are promising options for preventing the direct leakage of private video content, they remain vulnerable to side-channel attacks. We present Visor, a system that provides confidentiality for the user's video stream as well as the ML models in the presence of a compromised cloud platform and untrusted co-tenants. Visor executes video pipelines in a hybrid TEE that spans both the CPU and GPU. It protects the pipeline against side-channel attacks induced by data-dependent access patterns of video modules, and also addresses leakage in the CPU-GPU communication channel. Visor is up to $1000\times$ faster than na\"ive oblivious solutions, and its overheads relative to a non-oblivious baseline are limited to $2\times$--$6\times$.

arXiv:2006.10004공식 코드 · Papers with Code

Deeply Learned Spectral Total Variation Decomposition

Non-linear spectral decompositions of images based on one-homogeneous functionals such as total variation have gained considerable attention in the last few years. Due to their ability to extract spectral components corresponding to objects of different size and contrast, such decompositions enable filtering, feature transfer, image fusion and other applications. However, obtaining this decomposition involves solving multiple non-smooth optimisation problems and is therefore computationally highly intensive. In this paper, we present a neural network approximation of a non-linear spectral decomposition. We report up to four orders of magnitude ($\times 10,000$) speedup in processing of mega-pixel size images, compared to classical GPU implementations. Our proposed network, TVSpecNET, is able to implicitly learn the underlying PDE and, despite being entirely data driven, inherits invariances of the model based transform. To the best of our knowledge, this is the first approach towards learning a non-linear spectral decomposition of images. Not only do we gain a staggering computational advantage, but this approach can also be seen as a step towards studying neural networks that can decompose an image into spectral components defined by a user rather than a handcrafted functional.

arXiv:2006.09204

PlumeNet: Large-Scale Air Quality Forecasting Using A Convolutional LSTM Network

This paper presents an engine able to forecast jointly the concentrations of the main pollutants harming people's health: nitrogen dioxyde (NO2), ozone (O3) and particulate matter (PM2.5 and PM10, which are respectively the particles whose diameters are below 2.5 um and 10 um respectively). The forecasts are performed on a regular grid (the results presented in the paper are produced with a 0.5{\deg} resolution grid over Europe and the United States) with a neural network whose architecture includes convolutional LSTM blocks. The engine is fed with the most recent air quality monitoring stations measures available, weather forecasts as well as air quality physical and chemical model (AQPCM) outputs. The engine can be used to produce air quality forecasts with long time horizons, and the experiments presented in this paper show that the 4 days forecasts beat very significantly simple benchmarks. A valuable advantage of the engine is that it does not need much computing power: the forecasts can be built in a few minutes on a standard GPU. Thus, they can be updated very frequently, as soon as new air quality measures are available (generally every hour), which is not the case of AQPCMs traditionally used for air quality forecasting. The engine described in this paper relies on the same principles as a prediction engine deployed and used by Plume Labs in several products aiming at providing air quality data to individuals and businesses.

arXiv:2006.08861

GPU-accelerated Hierarchical Panoramic Image Feature Retrieval for Indoor Localization

Indoor localization has many applications, such as commercial Location Based Services (LBS), robotic navigation, and assistive navigation for the blind. This paper formulates the indoor localization problem into a multimedia retrieving problem by modeling visual landmarks with a panoramic image feature, and calculating a user's location via GPU- accelerated parallel retrieving algorithm. To solve the scene similarity problem, we apply a multi-images based retrieval strategy and a 2D aggregation method to estimate the final retrieval location. Experiments on a campus building real data demonstrate real-time responses (14fps) and robust localization.

arXiv:2006.09134공식 코드 · Papers with Code

AlphaGAN: Fully Differentiable Architecture Search for Generative Adversarial Networks

Generative Adversarial Networks (GANs) are formulated as minimax game problems, whereby generators attempt to approach real data distributions by virtue of adversarial learning against discriminators. The intrinsic problem complexity poses the challenge to enhance the performance of generative networks. In this work, we aim to boost model learning from the perspective of network architectures, by incorporating recent progress on automated architecture search into GANs. To this end, we propose a fully differentiable search framework for generative adversarial networks, dubbed alphaGAN. The searching process is formalized as solving a bi-level minimax optimization problem, in which the outer-level objective aims for seeking a suitable network architecture towards pure Nash Equilibrium conditioned on the generator and the discriminator network parameters optimized with a traditional GAN loss in the inner level. The entire optimization performs a first-order method by alternately minimizing the two-level objective in a fully differentiable manner, enabling architecture search to be completed in an enormous search space. Extensive experiments on CIFAR-10 and STL-10 datasets show that our algorithm can obtain high-performing architectures only with 3-GPU hours on a single GPU in the search space comprised of approximate 2 ? 1011 possible configurations. We also provide a comprehensive analysis on the behavior of the searching process and the properties of searched architectures, which would benefit further research on architectures for generative models. Pretrained models and codes are available at https://github.com/yuesongtian/AlphaGAN.

arXiv:2006.09049

Multi-Precision Policy Enforced Training (MuPPET): A precision-switching strategy for quantised fixed-point training of CNNs

Large-scale convolutional neural networks (CNNs) suffer from very long training times, spanning from hours to weeks, limiting the productivity and experimentation of deep learning practitioners. As networks grow in size and complexity, training time can be reduced through low-precision data representations and computations. However, in doing so the final accuracy suffers due to the problem of vanishing gradients. Existing state-of-the-art methods combat this issue by means of a mixed-precision approach utilising two different precision levels, FP32 (32-bit floating-point) and FP16/FP8 (16-/8-bit floating-point), leveraging the hardware support of recent GPU architectures for FP16 operations to obtain performance gains. This work pushes the boundary of quantised training by employing a multilevel optimisation approach that utilises multiple precisions including low-precision fixed-point representations. The novel training strategy, MuPPET, combines the use of multiple number representation regimes together with a precision-switching mechanism that decides at run time the transition point between precision regimes. Overall, the proposed strategy tailors the training process to the hardware-level capabilities of the target hardware architecture and yields improvements in training time and energy efficiency compared to state-of-the-art approaches. Applying MuPPET on the training of AlexNet, ResNet18 and GoogLeNet on ImageNet (ILSVRC12) and targeting an NVIDIA Turing GPU, MuPPET achieves the same accuracy as standard full-precision training with training-time speedup of up to 1.84$\times$ and an average speedup of 1.58$\times$ across the networks.

arXiv:2006.07644공식 코드 · Papers with Code

RoadNet-RT: High Throughput CNN Architecture and SoC Design for Real-Time Road Segmentation

In recent years, convolutional neural network has gained popularity in many engineering applications especially for computer vision. In order to achieve better performance, often more complex structures and advanced operations are incorporated into the neural networks, which results very long inference time. For time-critical tasks such as autonomous driving and virtual reality, real-time processing is fundamental. In order to reach real-time process speed, a light-weight, high-throughput CNN architecture namely RoadNet-RT is proposed for road segmentation in this paper. It achieves 90.33% MaxF score on test set of KITTI road segmentation task and 8 ms per frame when running on GTX 1080 GPU. Comparing to the state-of-the-art network, RoadNet-RT speeds up the inference time by a factor of 20 at the cost of only 6.2% accuracy loss. For hardware design optimization, several techniques such as depthwise separable convolution and non-uniformed kernel size convolution are customized designed to further reduce the processing time. The proposed CNN architecture has been successfully implemented on an FPGA ZCU102 MPSoC platform that achieves the computation capability of 83.05 GOPS. The system throughput reaches 327.9 frames per second with image size 1216x176.

arXiv:2006.07637

SE-MelGAN -- Speaker Agnostic Rapid Speech Enhancement

Recent advancement in Generative Adversarial Networks in speech synthesis domain[3],[2] have shown, that it's possible to train GANs [8] in a reliable manner for high quality coherent waveform generation from mel-spectograms. We propose that it is possible to transfer the MelGAN's [3] robustness in learning speech features to speech enhancement and noise reduction domain without any model modification tasks. Our proposed method generalizes over multi-speaker speech dataset and is able to robustly handle unseen background noises during the inference. Also, we show that by increasing the batch size for this particular approach not only yields better speech results, but generalizes over multi-speaker dataset easily and leads to faster convergence. Additionally, it outperforms previous state of the art GAN approach for speech enhancement SEGAN [5] in two domains: 1. quality ; 2. speed. Proposed method runs at more than 100x faster than realtime on GPU and more than 2x faster than real time on CPU without any hardware optimization tasks, right at the speed of MelGAN [3].

arXiv:2006.07871

GP3: A Sampling-based Analysis Framework for Gaussian Processes

Although machine learning is increasingly applied in control approaches, only few methods guarantee certifiable safety, which is necessary for real world applications. These approaches typically rely on well-understood learning algorithms, which allow formal theoretical analysis. Gaussian process regression is a prominent example among those methods, which attracts growing attention due to its strong Bayesian foundations. Even though many problems regarding the analysis of Gaussian processes have a similar structure, specific approaches are typically tailored for them individually, without strong focus on computational efficiency. Thereby, the practical applicability and performance of these approaches is limited. In order to overcome this issue, we propose a novel framework called GP3, general purpose computation on graphics processing units for Gaussian processes, which allows to solve many of the existing problems efficiently. By employing interval analysis, local Lipschitz constants are computed in order to extend properties verified on a grid to continuous state spaces. Since the computation is completely parallelizable, the computational benefits of GPU processing are exploited in combination with multi-resolution sampling in order to allow high resolution analysis.

arXiv:2006.08509공식 코드 · Papers with Code

APQ: Joint Search for Network Architecture, Pruning and Quantization Policy

We present APQ for efficient deep learning inference on resource-constrained hardware. Unlike previous methods that separately search the neural architecture, pruning policy, and quantization policy, we optimize them in a joint manner. To deal with the larger design space it brings, a promising approach is to train a quantization-aware accuracy predictor to quickly get the accuracy of the quantized model and feed it to the search engine to select the best fit. However, training this quantization-aware accuracy predictor requires collecting a large number of quantized <model, accuracy> pairs, which involves quantization-aware finetuning and thus is highly time-consuming. To tackle this challenge, we propose to transfer the knowledge from a full-precision (i.e., fp32) accuracy predictor to the quantization-aware (i.e., int8) accuracy predictor, which greatly improves the sample efficiency. Besides, collecting the dataset for the fp32 accuracy predictor only requires to evaluate neural networks without any training cost by sampling from a pretrained once-for-all network, which is highly efficient. Extensive experiments on ImageNet demonstrate the benefits of our joint optimization approach. With the same accuracy, APQ reduces the latency/energy by 2x/1.3x over MobileNetV2+HAQ. Compared to the separate optimization approach (ProxylessNAS+AMC+HAQ), APQ achieves 2.3% higher ImageNet accuracy while reducing orders of magnitude GPU hours and CO2 emission, pushing the frontier for green AI that is environmental-friendly. The code and video are publicly available.