This paper outlines a method used to successfully modify the Seam Carving algorithm so that it can be run to remove objects from videos in realtime. The successful approach was a combination of multiprocessing speedu...
详细信息
ISBN:
(纸本)9781450386975
This paper outlines a method used to successfully modify the Seam Carving algorithm so that it can be run to remove objects from videos in realtime. The successful approach was a combination of multiprocessing speedups and what we called the "Spiral Model". Resultant removal took about 35 milliseconds. The performant object recognition technique was not novel, but resulted in a total processing delay of less than.1 seconds and reasonably continuous video streaming while detecting and removing an object of interest. This paper will discuss the methods used and display the visual results of the approach.
In order to improve security measures across numerous domains, such as public safety, transportation, and critical infrastructure protection, the use of closed-circuit television (CCTV) systems for threat analysis has...
详细信息
In the modern world, reckless driving has been a major contributing factor in many road accidents. Various researches and studies have already been done on speed detection and published on the Internet. Proposed appro...
详细信息
We propose a system that uses depth information, which represents the distance from the sensor, instead of color information to do both measure human flow and protect privacy for low-end IoT devices. The system is des...
详细信息
ISBN:
(纸本)9798350300673
We propose a system that uses depth information, which represents the distance from the sensor, instead of color information to do both measure human flow and protect privacy for low-end IoT devices. The system is designed to detect the position and number of persons from depth information. Since administrative organizations or educational institutions have been reducing their budgets, this system should be implemented at as low a cost as possible. In order to realize human flow detection system on low-end Iot devices, we use low performance depth cameras as data acquisition device controlled by single board computers, such as Raspberry Pi. As one of our goals is all data processing are performed on single board computers, we adopt computational methods for the detection as possible as simple. The proposed method is based on the background subtraction method, which prepares a reference depth image and extracts moving regions from one frame extracted from the video depth image. Furthermore, we aim to achieve high-precision people flow measurement by combining the following three elements: segmentation of moving objects using edge information, identification of human areas using floor information, and human tracking using areas where people overlap in the direction of the time axis. Experiments were also conducted and evaluated in a real space using a program that implements the presented method.
video object detection (VID) is challenging because of the high variation of object appearance as well as the diverse deterioration in some frames. On the positive side, the detection in a certain frame of a video, co...
详细信息
ISBN:
(纸本)9781577358800
video object detection (VID) is challenging because of the high variation of object appearance as well as the diverse deterioration in some frames. On the positive side, the detection in a certain frame of a video, compared with that in a still image, can draw support from other frames. Hence, how to aggregate features across different frames is pivotal to VID problem. Most of existing aggregation algorithms are customized for two-stage detectors. However, these detectors are usually computationally expensive due to their two-stage nature. This work proposes a simple yet effective strategy to address the above concerns, which costs marginal overheads with significant gains in accuracy. Concretely, different from traditional two-stage pipeline, we select important regions after the one-stage detection to avoid processing massive low-quality candidates. Besides, we evaluate the relationship between a target frame and reference frames to guide the aggregation. We conduct extensive experiments and ablation studies to verify the efficacy of our design, and reveal its superiority over other state-of-the-art VID approaches in both effectiveness and efficiency. Our YOLOX-based model can achieve promising performance (e.g., 87.5% AP50 at over 30 FPS on the imageNet VID dataset on a single 2080Ti GPU), making it attractive for large-scale or real-time applications. The implementation is simple, we have made the demo codes and models available at https://***/YuHengsss/YOLOV.
The proceedings contain 12 papers. The topics discussed include: digital eye on endangered wildlife: crafting recognition datasets through semi-automated annotation;a 3D model information hiding algorithm based on reg...
ISBN:
(纸本)9798400708275
The proceedings contain 12 papers. The topics discussed include: digital eye on endangered wildlife: crafting recognition datasets through semi-automated annotation;a 3D model information hiding algorithm based on region of interest blocked histogram adaptive shift;smooth test for multivariate skew-elliptical distributions;single-image HDR reconstruction based on mask-aware convolution;quantitative analysis of artificial validation sets for Fourier domain CT reconstruction;a two-stage network based on edge information for visual anomaly detection;efficient real-time dense pedestrian detector based on improved YOLOv7-tiny;face makeup transfer based on generative adversarial network;visual quality assessment of HDR omnidirectional image system based on viewport feature learning;and advancements in video-based insect tracking: a bibliometric analysis to a short survey.
image interpolation algorithm is an important branch of imageprocessing research. When traditional interpolation is applied, it is easy to cause problems such as edge aliasing and details blurring. In order to solve ...
详细信息
ISBN:
(数字)9798350361445
ISBN:
(纸本)9798350361452
image interpolation algorithm is an important branch of imageprocessing research. When traditional interpolation is applied, it is easy to cause problems such as edge aliasing and details blurring. In order to solve these problems, this paper combines Canny edge detection to improve the interpolation algorithm, which has better interpolation effect when dealing with images with more details. At the same time, the increase of algorithm complexity makes the traditional implementation method difficult to meet the real-time requirements of videoimageprocessing. Field Programmable Gate Array (FPGA) has excellent parallel processing ability. This design constructs and implements a real-timevideoimage interpolation algorithm combined with edge operator on the platform.
Convolutional neural network inference on video data requires powerful hardware for real-timeprocessing. Given the inherent coherence across consecutive frames, large parts of a video typically change little. By skip...
详细信息
ISBN:
(数字)9781665469463
ISBN:
(纸本)9781665469463
Convolutional neural network inference on video data requires powerful hardware for real-timeprocessing. Given the inherent coherence across consecutive frames, large parts of a video typically change little. By skipping identical image regions and truncating insignificant pixel updates, computational redundancy can in theory be reduced significantly. However, these theoretical savings have been difficult to translate into practice, as sparse updates hamper computational consistency and memory access coherence;which are key for efficiency on real hardware. With DeltaCNN, we present a sparse convolutional neural network framework that enables sparse frame-by-frame updates to accelerate video inference in practice. We provide sparse implementations for all typical CNN layers and propagate sparse feature updates end-to-end - without accumulating errors over time. DeltaCNN is applicable to all convolutional neural networks without retraining. To the best of our knowledge, we are the first to significantly outperform the dense reference, cuDNN, in practical settings, achieving speedups of up to 7x with only marginal differences in accuracy. Our CUDA kernels and PyTorch extensions can be found at https://***/facebookresearch/DeltaCNN.
The proceedings contain 173 papers. The topics discussed include: advancements in PCB defect detection: an in-depth exploration of imageprocessing techniques;implementing an interactive microcontroller toy for childr...
ISBN:
(纸本)9798350386349
The proceedings contain 173 papers. The topics discussed include: advancements in PCB defect detection: an in-depth exploration of imageprocessing techniques;implementing an interactive microcontroller toy for children with autism;implementation of patrolling robots for apartments using cloud technology;leveraging facial analytics for enhanced crime prevention - integrating video surveillance and FaceNet algorithm;detection of realtime pothole system using edge detection;blockchain-enabled decentralized trust management and secure voting system;sensing the health of a computer network through estimation of its reliability based on sensors response data analysis and its impact on network management;enhanced security fencing system with geolocation tracking;and deep learning based face regions identification to accurately detect human emotions.
The system proposed in this paper aims to assist new nurses with the accurate and timely identification of medical instruments used in the operating room or in the central sterile supply department. An important aspec...
详细信息
ISBN:
(纸本)9798331532154;9798331532147
The system proposed in this paper aims to assist new nurses with the accurate and timely identification of medical instruments used in the operating room or in the central sterile supply department. An important aspect of surgical intervention flow is the efficient communication between the operators and the rest of the medical staff to ensure the proper procedure times. That also means the positioning in order of the medical instruments on the Mayo trays, which is often a problem for new nurses. In those situations, quick identification is needed;otherwise, stressful situations within the medical team can be induced. Our approach implements a new method for rapid detection of the instruments in real-time using a low-cost video camera and development motherboard, which embeds two machine learning algorithms, MobileNets2 and YOLOv5, by Transfer Learning The results show that high identification accuracies can be obtained. We also emphasize that many operating rooms that are equipped with adjustable lights and operating cameras with such solutions can also be implemented as an automatic step for surgery preparation. Moreover, this can also avoid possible contamination between sterile and non sterile environments.
暂无评论