Deep Convolutional Neural Networks (CNN) are the state-of-the-art performers for the object detection task. It is well known that object detection requires more computation and memory than image classification. In thi...
详细信息
ISBN:
(纸本)9781538607336
Deep Convolutional Neural Networks (CNN) are the state-of-the-art performers for the object detection task. It is well known that object detection requires more computation and memory than image classification. In this work, we propose LCDet, a fully-convolutional neural network for generic object detection that aims to work in embedded systems. We design and develop an end-to-end TensorFlow(TF)-based model. The detection works by a single forward pass through the network. Additionally, we employ 8-bit quantization on the learned weights. As a use case, we choose face detection and train the proposed model on images containing a varying number of faces of different sizes. We evaluate the face detection performance on publicly available dataset FDDB and Widerface. Our experimental results show that the proposed method achieves comparative accuracy comparing with state-of-the-art CNN-based face detection methods while reducing the model size by 3x and memory-BW by 3 - 4x comparing with one of the best real-time CNN-based object detector YOLO [23]. Our 8-bit fixed-point TF-model provides additional 4x memory reduction while keeping the accuracy nearly as good as the floating point model and achieves 20x performance gain compared to the floating point model. Thus the proposed model is amenable for embedded implementations and is generic to be extended to any number of categories of objects.
We describe a new Objective Video Quality Assessment (VQA) metric, consisting of a method based on spatio-temporal saliency to model human visual perception of quality. Accurate measurement of video quality is an impo...
详细信息
In this paper, hardware implementation of edge detection at real time video signals using Sobel, Robert, Prewitt and Laplacian filters based on FPGA is explained. Besides, filters are compared in many ways. Edge detec...
详细信息
ISBN:
(纸本)9781509064953
In this paper, hardware implementation of edge detection at real time video signals using Sobel, Robert, Prewitt and Laplacian filters based on FPGA is explained. Besides, filters are compared in many ways. Edge detection is an elemantary and fundamental tool for image segmentation and feature extraction. Very high speed hardware like FPGA's are used to implement the image and video processingalgorithms for improving the performance of processingsystems. algorithms are implemented on the Xilinx Zynq 7000. The video input signals come from a laptop's HDMI interface to FPGA in order to filter and the detected edges are displayed on a HDMI display screen.
The proceedings contain 30 papers. The special focus in this conference is on Applications of Machine Learning, Cloud Computing, Transportation, Multi-Robot systems and Uncertain systems. The topics include: Maximum l...
ISBN:
(纸本)9783319489438
The proceedings contain 30 papers. The special focus in this conference is on Applications of Machine Learning, Cloud Computing, Transportation, Multi-Robot systems and Uncertain systems. The topics include: Maximum likelihood estimation and optimal coordinates;relation recognition problems and algebraic approach to their solution;prediction of power load demand using modified dynamic weighted majority method;estimating cluster population;evaluation of particle swarm optimisation for medical image segmentation;automated processing of micro-ct scans using descriptor-based registration of 3d images;topic modeling based on frequent sequences graphs;Gaussian process regression with categorical inputs for predicting the blood glucose level;automated information extraction and classification of matrix-based questionnaire data;evaluating raft in docker on kubernetes;performance evaluation of MPTCP transmission of large data objects in computing cloud;a decentralized system for load balancing of containerized microservices in the cloud;layered reconfigurable architecture for autonomous cooperative UAV computing systems;a practical verification of protocol and data format negotiation methods in ComSS platform;reactive dynamic assignment for a bi-dimensional traffic flow model;comparing signal setting design methods through emission and fuel consumption performance indicators;GSOM traffic flow models for networks with information;designing mass-customized service subject to public grid-like network constraints;sensing feedback for the control of multi-joint prosthetic hand;spatio-temporal clustering and forecasting method for free-floating bike sharing systems;comparison of algorithms for constrained multi-robot task allocation and a joint problem of track closure planning and train run rescheduling with detours.
Camera sensors can only capture a limited range of luminance simultaneously, and in order to create high dynamic range (HDR) images a set of different exposures are typically combined. In this paper we address the pro...
详细信息
Camera sensors can only capture a limited range of luminance simultaneously, and in order to create high dynamic range (HDR) images a set of different exposures are typically combined. In this paper we address the problem of predicting information that have been lost in saturated image areas, in order to enable HDR reconstruction from a single exposure. We show that this problem is well-suited for deep learning algorithms, and propose a deep convolutional neural network (CNN) that is specifically designed taking into account the challenges in predicting HDR values. To train the CNN we gather a large dataset of HDR images, which we augment by simulating sensor saturation for a range of cameras. To further boost robustness, we pre-train the CNN on a simulated HDR dataset created from a subset of the MIT Places database. We demonstrate that our approach can reconstruct high-resolution visually convincing HDR results in a wide range of situations, and that it generalizes well to reconstruction of images captured with arbitrary and low-end cameras that use unknown camera response functions and post-processing. Furthermore, we compare to existing methods for HDR expansion, and show high quality results also for image based lighting. Finally, we evaluate the results in a subjective experiment performed on an HDR display. This shows that the reconstructed HDR images are visually convincing, with large improvements as compared to existing methods.
The paper proposes a modification of the pyramid method for constructing algorithms for the difference solution of the d'Alembert equation on a graphics processor in the event of a shortage of video memory. The au...
详细信息
The paper proposes a modification of the pyramid method for constructing algorithms for the difference solution of the d'Alembert equation on a graphics processor in the event of a shortage of video memory. The authors demonstrate the effectiveness of the method on the practical example of dividing the grid area into two sub domains. Acceleration reaches the characteristic for the case of a domain entirely located in the video memory. In the article investigated the effectiveness of using the author's approach depending on the height of the pyramid and showed the boundaries of applicability of the proposed modification.
The proceedings contain 50 papers. The topics discussed include: on high-precision chessboard detection on static scene videos from mobile eye-tracking devices;towards a mobile system for hypertensive outpatients'...
ISBN:
(纸本)9789526865324
The proceedings contain 50 papers. The topics discussed include: on high-precision chessboard detection on static scene videos from mobile eye-tracking devices;towards a mobile system for hypertensive outpatients' treatment adherence improvement;power distribution unit (PDU) for a distributed computing network;analysis of the functioning of a multi-domain transport software-defined network with controlled optical layer;not-holding connections as a characteristic feature of dynamic interactions of elements of technical systems;an algorithm for building an enterprise network topology using widespread data sources;a web of things approach for indoor position monitoring of elderly and impaired people;modular industrial equipment in cyber-physical production system: architecture and integration;a hyper heuristic algorithm for scheduling of fog networks;digital image watermarking using dwt basis matrices;2D SLAM quality evaluation methods;revealing of entities interconnections in system dynamics modelling process by applying multimodal data analysis paradigm;fast and modular regularized topic modelling;and modern methods and algorithms in digital processing of endoscopic images.
Demand of multimedia growth, contributes to insufficient bandwidth of network and memory storage device. Therefore data compression is more required for reducing data redundancy to save more hardware space and transmi...
详细信息
ISBN:
(纸本)9781509047611
Demand of multimedia growth, contributes to insufficient bandwidth of network and memory storage device. Therefore data compression is more required for reducing data redundancy to save more hardware space and transmission bandwidth. image compression is one of the main research in the field of imageprocessing. Many techniques are given for image compression. Some of which are discussed in this paper. This paper discusses k-means clustering, 2D-DWT and fuzzy logic based image compression.
Now-a-days roads are monitored by CCTV cameras passively i.e. knowledge is not generated by the CCTV cameras. Human resource constraint i.e. there is very limited posting of the traffic police staff at road crossings ...
详细信息
ISBN:
(纸本)9781538611869
Now-a-days roads are monitored by CCTV cameras passively i.e. knowledge is not generated by the CCTV cameras. Human resource constraint i.e. there is very limited posting of the traffic police staff at road crossings and highway's. In many cases, accidents are reported due to negligence by the road users. Adding to this, even the majority of humans surrounding the accident spot/scene are busy clicking photos and videos unknown of the fact that their little negligence could cost a LIFE. The footage is only seen by the authorities only after a problem has been occurred so as to find the root cause of it. To overcome this, we propose an intelligent system that can make use of the existing CCTV cameras. The proposed system captures video stream, computes the input and the system alerts are generated in real time, which means no additional sensors would be required. U sing camera itself, the plan is to detect accidents in real time and also send alerts to ambulances or medical services; so that appropriate resources needed for saving lives are available in time. Other targets are: to detect vehicles which break the speed limit rules and relay this information to a concerned authority; to monitor restricted roads areas and traffic signal violations, and plan to detect number plates of the violators and find them. Along with this the proposed system will also classify the type of vehicles that run on road, so that further construction of roads can be tailored as per the expected vehicles for smoother traffic flow. In this study we will be taking help of algorithms such as Background Subtraction, Morphological transformations and many other basic concepts to perform the tasks mentioned above. By using technology at its best, the aim is to integrate various factors into one smart system. The project will benefit to reduce cost of road surveillance system and complete automation of road surveillance system.
This paper is devoted to present technique of the use of imageprocessing for lab-on-a-chip techniques. algorithms and methods for cell detecting, obtaining their parameters and multiparametric cell tracking in lab-on...
详细信息
ISBN:
(纸本)9781509064861
This paper is devoted to present technique of the use of imageprocessing for lab-on-a-chip techniques. algorithms and methods for cell detecting, obtaining their parameters and multiparametric cell tracking in lab-on-a-chip were presented and discussed from the point of real-time detection.
暂无评论