The proceedings contain 17 papers. The special focus in this conference is on Computer Vision, Graphics, and imageprocessing. The topics include: The Ikshana Hypothesis of Human Scene Understanding;worst-Case Adversa...
ISBN:
(纸本)9789811941351
The proceedings contain 17 papers. The special focus in this conference is on Computer Vision, Graphics, and imageprocessing. The topics include: The Ikshana Hypothesis of Human Scene Understanding;worst-Case Adversarial Perturbation and Effect of Feature Normalization on Max-Margin Multi-label Classifiers;Catch Me if You Can: A Novel Task for Detection of Covert Geo-Locations (CGL);MATIC: Memory-Guided Adaptive Transformer for image Captioning;Semantic Map Injected GAN Training for image-to-image Translation;textGen3D: A real-time 3D-Mesh Generation with Intersecting Contours for Text;comparative Analysis of Neural Architecture Search Methods for Classification of Cultural Heritage Sites;heritage Representation of Kashi Vishweshwar Temple at Kalabgoor, Telangana with Augmented reality Application Using Photogrammetry;augmented Data as an Auxiliary Plug-In Toward Categorization of Crowdsourced Heritage Data;evolution of Bagbazar Street Through Visibility Graph Analysis (1746–2020);mapping Archaeological Remains of 14th Century Fort of Jahanpanah Using Geospatial Analysis;spatial Analysis and 3d Mapping Historic Landscapes—Implications of Adopting an Integrated Approach in Simulation and Visualization of Landscapes;HSADML: Hyper-Sphere Angular Deep Metric Based Learning for Brain Tumor Classification;model Compression Based Lightweight Online Signature Verification Framework.
This study takes the theme of underwater shrimp digital image segmentation using edge detection methods on fog networks, the problem raised in this study is to compare the accuracy of the shrimp digital image segmenta...
详细信息
Until recently, physicians conducted several diagnostic gait & posture tests with simple time measurements, using a stopwatch. Nowadays, with the help of AI, such tests have been automated, i.e., performed without...
详细信息
ISBN:
(纸本)9798400713170
Until recently, physicians conducted several diagnostic gait & posture tests with simple time measurements, using a stopwatch. Nowadays, with the help of AI, such tests have been automated, i.e., performed without the presence of a physician. Application stand2squatAI_biorig is an example of a realtime measurement method. In the current research we conducted 26 stand to squat tests, to estimate the precision of its measurements by comparing it against video recording and stopwatch timing. The value of the application under study proved significant, since it outperformed stopwatch timing. Furthermore, it performed on par with the video recording method which required significant post-processing for manually time-stamping the stances.
Traditional network traffic detection methods based on machine learning have been widely used, but the accuracy of identifying and detecting abnormal traffic is not high. Compared with machine learning methods, deep l...
详细信息
ISBN:
(数字)9798331533694
ISBN:
(纸本)9798331533700
Traditional network traffic detection methods based on machine learning have been widely used, but the accuracy of identifying and detecting abnormal traffic is not high. Compared with machine learning methods, deep learning has made significant progress in fields such as text processing, video generation, and other fields, bringing new ideas to abnormal traffic detection. This paper adopts an abnormal traffic classification method based on multi-pooling cascaded convolutional neural network (MPCNN). The accuracy of extensive experiments is 0.9914, the precision is 0.9984, and the F1 score is 0.9962. These experiments achieved the company's goal of accuracy greater than 0.98. It is used for real-time monitoring of daily network traffic to recognize abnormal traffic virtually, alert topic timely, and prevent potential network threats.
Edge video Analytics (EVA) has become a major application of pervasive computing, enabling real-time visual processing. EVA pipelines, composed of deep neural networks (DNNs), typically demand efficient inference serv...
详细信息
ISBN:
(数字)9798331535513
ISBN:
(纸本)9798331535520
Edge video Analytics (EVA) has become a major application of pervasive computing, enabling real-time visual processing. EVA pipelines, composed of deep neural networks (DNNs), typically demand efficient inference serving under stringent latency requirements, which is challenging due to the dynamic Edge environments (e.g., workload variability and network instability). Moreover, EVA pipelines face significant resource contention due to resource (e.g., GPU) constraints at the Edge. In this paper, we introduce OctopInf, a novel resource-efficient and workload-aware inference serving system designed for real-time EVA. OctopInf tackles the unique challenges of dynamic edge environments through fine-grained resource allocation, adaptive batching, and workload balancing between edge devices and servers. Furthermore, we propose a spatiotemporal scheduling algorithm that optimizes the co-location of inference tasks on GPUs, improving performance and ensuring service-level objectives (SLOs) compliance. Extensive evaluations on a real-world testbed demonstrate the effectiveness of our approach. It achieves an effective throughput increase of up to 10× compared to the baselines and shows better robustness in challenging scenarios. OctopInf can be used for any DNN-based EVA inference task with minimal adaptation and is available at https://***/tungngreen/PipelineScheduler.
This paper is devoted to solving the issue of improving the quality of video content (digital photos and videos) in real-time apps. This work aims to improve the efficiency of enhancing complex images through adaptive...
详细信息
ISBN:
(纸本)9781665438940
This paper is devoted to solving the issue of improving the quality of video content (digital photos and videos) in real-time apps. This work aims to improve the efficiency of enhancing complex images through adaptive power-law intensity transformations in automatic mode. In this paper, a new approach to image enhancement in an automatic mode based on adaptive power-law image intensity transformation (APLIT) was proposed. The APLIT approach allows directly calculating the value of the exponent of the gamma correction for the set thresholds and reducing the computational costs compared with the traditional technique of power-law transformation. Based on APLIT, we have proposed two new techniques of adaptive power-law transformation. The proposed APLIT based techniques show results close to the traditional gamma correction but have lower computational costs. The proposed techniques are efficient, easy to implement, computationally low-cost, and are designed to normalize and enhance video in real-time apps.
While conventional image compression techniques are optimized for human visual perception, the rise of machine learning techniques has led to the emergence of image compression methods tailored for machine vision task...
详细信息
We propose a novel Parallel Monte Carlo tree search with Batched Simulations (PMBS) algorithm for accelerating long-horizon, episodic robotic planning tasks. Monte Carlo tree search (MCTS) is an effective heuristic se...
详细信息
ISBN:
(数字)9781665479271
ISBN:
(纸本)9781665479271
We propose a novel Parallel Monte Carlo tree search with Batched Simulations (PMBS) algorithm for accelerating long-horizon, episodic robotic planning tasks. Monte Carlo tree search (MCTS) is an effective heuristic search algorithm for solving episodic decision-making problems whose underlying search spaces are expansive. Leveraging a GPU-based large-scale simulator, PMBS introduces massive parallelism into MCTS for solving planning tasks through the batched execution of a large number of concurrent simulations, which allows for more efficient and accurate evaluations of the expected cost-to-go over large action spaces. When applied to the challenging manipulation tasks of object retrieval from clutter, PMBS achieves a speedup of over 30x with an improved solution quality, in comparison to a serial MCTS implementation. We show that PMBS can be directly applied to real robot hardware with negligible sim-to-real differences. Supplementary material, including video, can be found at https://***/arc-l/pmbs.
The paper introduces an invisible image watermarking approach using a binary image as a watermark in combination with the Hermite transform. By using the secret key, the binary watermark is separated into several part...
详细信息
ISBN:
(数字)9789532901382
ISBN:
(纸本)9798350354614
The paper introduces an invisible image watermarking approach using a binary image as a watermark in combination with the Hermite transform. By using the secret key, the binary watermark is separated into several partial binary images before being embedded into the transform domain coefficients. Embedding is done into several bit planes of the Hermite coefficients to ensure invisibility and robustness. The Hermite transform proves to be efficient in diverse applications and outperforms traditional approaches, especially in block-based imageprocessing. Namely, it is shown that the Hermite transform can be used for image deblocking to mitigate the impact of ringing artifacts and blurring introduced by using the Fourier or discrete cosine transform-based approaches. Our focus is on defining a low-complexity architecture for real-time Hermite-based image watermarking. Therefore, the procedure is adopted to work in a vector manner to facilitate the hardware implementation. The schematic depiction of the design is provided and the complexity is discussed.
暂无评论