This work is part of a research project carried out during the COVID-19 pandemic, involving the design and realization of an autonomous mobile hospital robot. Many real-world robotic tasks suffer from the critical cha...
详细信息
images captured in poor lighting conditions (such haze, fog, mist, or smog) have a lower level of visibility because air particles deflect light. Single picture dehazing techniques can restore clarity to a single hazy...
详细信息
Machine learning offers the potential to enhance real-time image analysis in surgical operations. This paper presents results from the implementation of machine learning algorithms targeted for an intelligent image pr...
详细信息
Obtaining 3D surface information and physical material information of an object from images is an essential research prospect in computer vision and computer graphics. image-based 3D reconstruction is to extract the 3...
详细信息
ISBN:
(数字)9789811924484
ISBN:
(纸本)9789811924484;9789811924477
Obtaining 3D surface information and physical material information of an object from images is an essential research prospect in computer vision and computer graphics. image-based 3D reconstruction is to extract the 3D depth information of the scene and objects from single or multiple images through specific algorithms to reconstruct the 3D model of objects or locations with robust realism, which has fast reconstruction speed, simple equipment, realistic effect, and minor technical data, which can better realize the virtualization of natural objects.
Equivariances provide useful inductive biases in neural network modeling, with the translation equivariance of convolutional neural networks being a canonical example. Equivariances can be embedded in architectures th...
详细信息
ISBN:
(纸本)9781713871088
Equivariances provide useful inductive biases in neural network modeling, with the translation equivariance of convolutional neural networks being a canonical example. Equivariances can be embedded in architectures through weight-sharing and place symmetry constraints on the functions a neural network can represent. The type of symmetry is typically fixed and has to be chosen in advance. Although some tasks are inherently equivariant, many tasks do not strictly follow such symmetries. In such cases, equivariance constraints can be overly restrictive. In this work, we propose a parameter-efficient relaxation of equivariance that can effectively interpolate between a (i) non-equivariant linear product, (ii) a strict-equivariant convolution, and (iii) a strictly-invariant mapping. The proposed parameterisation can be thought of as a building block to allow adjustable symmetry structure in neural networks. In addition, we demonstrate that the amount of equivariance can be learned from the training data using backpropagation. Gradient-based learning of equivariance achieves similar or improved performance compared to the best value found by cross-validation and outperforms baselines with partial or strict equivariance on CIFAR-10 and CIFAR-100 image classification tasks.
The multispectral imaging which are used for remote sensing imaging has a large amount of data, so this paper proposed a deep learningmethod which is based on sliced convolutional LSTM for multispectral image compress...
详细信息
ISBN:
(纸本)9789811903908;9789811903892
The multispectral imaging which are used for remote sensing imaging has a large amount of data, so this paper proposed a deep learningmethod which is based on sliced convolutional LSTM for multispectral image compression. Compared with other algorithms, the proposed algorithm further compresses the multispectral images by considering the similarity between the spectra and removing the inter-spectral redundancy. The proposed algorithm is based on end to end framework which is consist of encoder, decoder, entropy coding and quantizer. In experiments, the PSNR of proposed model is compared with that of JPEG2000 to evaluate the performance of our algorithm at several different bit rates.
The paper considers the problem of clustering pixels of a color raster image. The task is to compare the effectiveness of three different clustering methods: k-means, DBSCAN, agglomerative clustering. The k-means and ...
详细信息
ISBN:
(数字)9798331532178
ISBN:
(纸本)9798331532185
The paper considers the problem of clustering pixels of a color raster image. The task is to compare the effectiveness of three different clustering methods: k-means, DBSCAN, agglomerative clustering. The k-means and agglomerative clustering algorithms consider different numbers of clusters: 2, 5, 10, 15, 20. The effectiveness is assessed using the SSIM metric and visual analysis of the resulting images.
Architectural designers and technologists are able to make an assessment on buildability, thermal and hygrothermal performance of design details. To process drawings, human vision segments, classifies and distinguishe...
详细信息
ISBN:
(纸本)9789811662690;9789811662683
Architectural designers and technologists are able to make an assessment on buildability, thermal and hygrothermal performance of design details. To process drawings, human vision segments, classifies and distinguishes the drawing objects on the basis of their knowledge. With the rapid advancement of Artificial Intelligence methods, vast opportunities become available for performing tasks that used to require human intelligence or assistance by humans. imageprocessing and analysis is one of these tasks that consists of the manipulation of images using algorithms. There are various applications in different fields, and the use of it is increasing exponentially. This paper explores the use of imageprocessing in identifying building materials in order to check compliance with building regulations and identify anomalies. In this paper, an encoder-decoder based deep convolutional neural network (DRU-net) for image segmentation is applied on architectural images to segment various materials including insulations, bricks and concrete in the conceptual development phase. An experimental analysis is performed on numerous detail drawings and an evaluation is made by mathematical models.
This paper puts forward a novel coalition formation framework to enable teams of Unmanned Aerial Vehicles (UAVs) to assist in victims localization and rescue in the case of post-avalanche events. Our framework consist...
详细信息
ISBN:
(纸本)9781450395977
This paper puts forward a novel coalition formation framework to enable teams of Unmanned Aerial Vehicles (UAVs) to assist in victims localization and rescue in the case of post-avalanche events. Our framework consists of multiple components. First, a novel initial UAVs placement algorithm (adapting a known Brain-Storming Optimization algorithm to our setting);second, a novel coalition structure generation protocol that allows for the "online" calculation of coalition values that will eventually guide the rescue effort;and third, a simple but effective opinion aggregation protocol, that can be used to prioritize certain rescue operations in the event of "ambiguous" findings. Moreover, for the recognition task itself we employ two image recognition algorithms, used for the first time for the analysis of post-avalanche Search & Rescue Operations images. Our experimental evaluation verifies the applicability and effectiveness of our framework and its individual components.
Video-quality measurement is a critical task in video processing. Nowadays, many implementations of new encoding standards - such as AV1, VVC, and LCEVC - use deep-learning-based decoding algorithms with perceptual me...
详细信息
ISBN:
(纸本)9781713871088
Video-quality measurement is a critical task in video processing. Nowadays, many implementations of new encoding standards - such as AV1, VVC, and LCEVC - use deep-learning-based decoding algorithms with perceptual metrics that serve as optimization objectives. But investigations of the performance of modern video- and image-quality metrics commonly employ videos compressed using older standards, such as AVC. In this paper, we present a new benchmark for video-quality metrics that evaluates video compression. It is based on a new dataset consisting of about 2,500 streams encoded using different standards, including AVC, HEVC, AV1, VP9, and VVC. Subjective scores were collected using crowdsourced pairwise comparisons. The list of evaluated metrics includes recent ones based on machine learning and neural networks. The results demonstrate that new no-reference metrics exhibit high correlation with subjective quality and approach the capability of top full-reference metrics.
暂无评论