arXiv subjects

Alexey Artemov

Publications and source records attributed to Alexey Artemov.

31 records · Page 2Linked to original sources

arXiv2020

Making DensePose fast and light

DensePose estimation task is a significant step forward for enhancing user experience computer vision applications ranging from augmented reality to cloth fitting. Existing neural network models capable of solving this task are heavily parameterized and a long way from being transferred to an embedded or mobile device. To enable Dense Pose inference on the end device with current models, one needs to support an expensive server-side infrastructure and have a stable internet connection. To make things worse, mobile and embedded devices do not always have a powerful GPU inside. In this work, we target the problem of redesigning the DensePose R-CNN model's architecture so that the final network retains most of its accuracy but becomes more light-weight and fast. To achieve that, we tested and incorporated many deep learning innovations from recent years, specifically performing an ablation study on 23 efficient backbone architectures, multiple two-stage detection pipeline modifications, and custom model quantization methods. As a result, we achieved $17\times$ model size reduction and $2\times$ latency improvement compared to the baseline model.

cs.CV↗

arXiv2020

Latent Video Transformer

The video generation task can be formulated as a prediction of future video frames given some past frames. Recent generative models for videos face the problem of high computational requirements. Some models require up to 512 Tensor Processing Units for parallel training. In this work, we address this problem via modeling the dynamics in a latent space. After the transformation of frames into the latent space, our model predicts latent representation for the next frames in an autoregressive manner. We demonstrate the performance of our approach on BAIR Robot Pushing and Kinetics-600 datasets. The approach tends to reduce requirements to 8 Graphical Processing Units for training the models while maintaining comparable generation quality.

cs.CV↗

arXiv2019

Latent-Space Laplacian Pyramids for Adversarial Representation Learning with 3D Point Clouds

Constructing high-quality generative models for 3D shapes is a fundamental task in computer vision with diverse applications in geometry processing, engineering, and design. Despite the recent progress in deep generative modelling, synthesis of finely detailed 3D surfaces, such as high-resolution point clouds, from scratch has not been achieved with existing approaches. In this work, we propose to employ the latent-space Laplacian pyramid representation within a hierarchical generative model for 3D point clouds. We combine the recently proposed latent-space GAN and Laplacian GAN architectures to form a multi-scale model capable of generating 3D point clouds at increasing levels of detail. Our evaluation demonstrates that our model outperforms the existing generative models for 3D point clouds.

cs.CV↗

arXiv2019

Weakly Supervised Fine Tuning Approach for Brain Tumor Segmentation Problem

Segmentation of tumors in brain MRI images is a challenging task, where most recent methods demand large volumes of data with pixel-level annotations, which are generally costly to obtain. In contrast, image-level annotations, where only the presence of lesion is marked, are generally cheap, generated in far larger volumes compared to pixel-level labels, and contain less labeling noise. In the context of brain tumor segmentation, both pixel-level and image-level annotations are commonly available; thus, a natural question arises whether a segmentation procedure could take advantage of both. In the present work we: 1) propose a learning-based framework that allows simultaneous usage of both pixel- and image-level annotations in MRI images to learn a segmentation model for brain tumor; 2) study the influence of comparative amounts of pixel- and image-level annotations on the quality of brain tumor segmentation; 3) compare our approach to the traditional fully-supervised approach and show that the performance of our method in terms of segmentation quality may be competitive.

eess.IV↗

arXiv2019

Perceptual deep depth super-resolution

RGBD images, combining high-resolution color and lower-resolution depth from various types of depth sensors, are increasingly common. One can significantly improve the resolution of depth maps by taking advantage of color information; deep learning methods make combining color and depth information particularly easy. However, fusing these two sources of data may lead to a variety of artifacts. If depth maps are used to reconstruct 3D shapes, e.g., for virtual reality applications, the visual quality of upsampled images is particularly important. The main idea of our approach is to measure the quality of depth map upsampling using renderings of resulting 3D surfaces. We demonstrate that a simple visual appearance-based loss, when used with either a trained CNN or simply a deep prior, yields significantly improved 3D shapes, as measured by a number of existing perceptual metrics. We compare this approach with a number of existing optimization and learning-based techniques.

cs.CV↗

arXiv2019

Learning to Approximate Directional Fields Defined over 2D Planes

Reconstruction of directional fields is a need in many geometry processing tasks, such as image tracing, extraction of 3D geometric features, and finding principal surface directions. A common approach to the construction of directional fields from data relies on complex optimization procedures, which are usually poorly formalizable, require a considerable computational effort, and do not transfer across applications. In this work, we propose a deep learning-based approach and study the expressive power and generalization ability.

cs.CV↗

arXiv2019

Procedural Synthesis of Remote Sensing Images for Robust Change Detection with Neural Networks

Data-driven methods such as convolutional neural networks (CNNs) are known to deliver state-of-the-art performance on image recognition tasks when the training data are abundant. However, in some instances, such as change detection in remote sensing images, annotated data cannot be obtained in sufficient quantities. In this work, we propose a simple and efficient method for creating realistic targeted synthetic datasets in the remote sensing domain, leveraging the opportunities offered by game development engines. We provide a description of the pipeline for procedural geometry generation and rendering as well as an evaluation of the efficiency of produced datasets in a change detection scenario. Our evaluations demonstrate that our pipeline helps to improve the performance and convergence of deep learning models when the amount of real-world data is severely limited.

cs.CV↗

arXiv2019

Monocular 3D Object Detection via Geometric Reasoning on Keypoints

Monocular 3D object detection is well-known to be a challenging vision task due to the loss of depth information; attempts to recover depth using separate image-only approaches lead to unstable and noisy depth estimates, harming 3D detections. In this paper, we propose a novel keypoint-based approach for 3D object detection and localization from a single RGB image. We build our multi-branch model around 2D keypoint detection in images and complement it with a conceptually simple geometric reasoning method. Our network performs in an end-to-end manner, simultaneously and interdependently estimating 2D characteristics, such as 2D bounding boxes, keypoints, and orientation, along with full 3D pose in the scene. We fuse the outputs of distinct branches, applying a reprojection consistency loss during training. The experimental evaluation on the challenging KITTI dataset benchmark demonstrates that our network achieves state-of-the-art results among other monocular 3D detectors.

cs.CV↗

arXiv2019

ABC: A Big CAD Model Dataset For Geometric Deep Learning

We introduce ABC-Dataset, a collection of one million Computer-Aided Design (CAD) models for research of geometric deep learning methods and applications. Each model is a collection of explicitly parametrized curves and surfaces, providing ground truth for differential quantities, patch segmentation, geometric feature detection, and shape reconstruction. Sampling the parametric descriptions of surfaces and curves allows generating data in different formats and resolutions, enabling fair comparisons for a wide range of geometric learning algorithms. As a use case for our dataset, we perform a large-scale benchmark for estimation of surface normals, comparing existing data driven methods and evaluating their performance against both the ground truth and traditional normal estimation methods.

cs.GR↗

arXiv2018

Machine Learning pipeline for discovering neuroimaging-based biomarkers in neurology and psychiatry

We consider a problem of diagnostic pattern recognition/classification from neuroimaging data. We propose a common data analysis pipeline for neuroimaging-based diagnostic classification problems using various ML algorithms and processing toolboxes for brain imaging. We illustrate the pipeline application by discovering new biomarkers for diagnostics of epilepsy and depression based on clinical and MRI/fMRI data for patients and healthy volunteers.

cs.CV↗

arXiv2018

fMRI: preprocessing, classification and pattern recognition

As machine learning continues to gain momentum in the neuroscience community, we witness the emergence of novel applications such as diagnostics, characterization, and treatment outcome prediction for psychiatric and neurological disorders, for instance, epilepsy and depression. Systematic research into these mental disorders increasingly involves drawing clinical conclusions on the basis of data-driven approaches; to this end, structural and functional neuroimaging serve as key source modalities. Identification of informative neuroimaging markers requires establishing a comprehensive preparation pipeline for data which may be severely corrupted by artifactual signal fluctuations. In this work, we review a large body of literature to provide ample evidence for the advantages of pattern recognition approaches in clinical applications, overview advanced graph-based pattern recognition approaches, and propose a noise-aware neuroimaging data processing pipeline. To demonstrate the effectiveness of our approach, we provide results from a pilot study, which show a significant improvement in classification accuracy, indicating a promising research direction.

cs.CV↗

arXiv2016

Detecting Performance Degradation of Software-Intensive Systems in the Presence of Trends and Long-Range Dependence

As contemporary software-intensive systems reach increasingly large scale, it is imperative that failure detection schemes be developed to help prevent costly system downtimes. A promising direction towards the construction of such schemes is the exploitation of easily available measurements of system performance characteristics such as average number of processed requests and queue size per unit of time. In this work, we investigate a holistic methodology for detection of abrupt changes in time series data in the presence of quasi-seasonal trends and long-range dependence with a focus on failure detection in computer systems. We propose a trend estimation method enjoying optimality properties in the presence of long-range dependent noise to estimate what is considered "normal" system behaviour. To detect change-points and anomalies, we develop an approach based on the ensembles of "weak" detectors. We demonstrate the performance of the proposed change-point detection scheme using an artificial dataset, the publicly available Abilene dataset as well as the proprietary geoinformation system dataset.

stat.AP↗

arXiv2015

Event Index - an LHCb Event Search System

During LHC Run 1, the LHCb experiment recorded around $10^{11}$ collision events. This paper describes Event Index - an event search system. Its primary function is to quickly select subsets of events from a combination of conditions, such as the estimated decay channel or number of hits in a subdetector. Event Index is essentially Apache Lucene optimized for read-only indexes distributed over independent shards on independent nodes.

hep-ex↗