Telemedicine is a promising direction in the development of medical technologies for the interaction of patients with doctors at a distance. In this paper, we consider the use of telemedicine technologies for the deve...
详细信息
ISBN:
(纸本)9783030308599;9783030308582
Telemedicine is a promising direction in the development of medical technologies for the interaction of patients with doctors at a distance. In this paper, we consider the use of telemedicine technologies for the development of smart medical autonomous technology. An example of a smart medical autonomous distributed system for diagnostics is also discussed. To develop this system for medical image analysis we review several processing methods and machine learning algorithms. Some examples of medical system processing results are presented.
Reasoning is an important ability that we learn from a very early age. Yet, reasoning is extremely hard for algorithms. Despite impressive recent progress that has been reported on tasks that necessitate reasoning, su...
详细信息
Reasoning is an important ability that we learn from a very early age. Yet, reasoning is extremely hard for algorithms. Despite impressive recent progress that has been reported on tasks that necessitate reasoning, such as visual question answering and visual dialog, models often exploit biases in datasets. To develop models with better reasoning abilities, recently, the new visual commonsense reasoning (VCR) task has been introduced. Not only do models have to answer questions, but also do they have to provide a reason for the given answer. The proposed baseline achieved compelling results, leveraging a meticulously designed model composed of LSTM modules and attention nets. Here we show that a much simpler model obtained by ablating and pruning the existing intricate baseline can perform better with half the number of trainable parameters. By associating visual features with attribute information and better text to image grounding, we obtain further improvements for our simpler & effective baseline, TAB-VCR. We show that this approach results in a 5.3%, 4.4% and 6.5% absolute improvement over the previous state-of-the-art [103] on question answering, answer justification and holistic VCR.
Modern remote sensing (RS) image application systems often distribute imageprocessing tasks among multiple data centers and then gather the processed images from each center to efficiently synthesize the final produc...
详细信息
ISBN:
(数字)9781728182544
ISBN:
(纸本)9781728182551
Modern remote sensing (RS) image application systems often distribute imageprocessing tasks among multiple data centers and then gather the processed images from each center to efficiently synthesize the final product. In this paper, we exploit the edge-cloud architecture to design and implement a novel RS image service system, called RS-pCloud, which leverages the Peer-to-Peer (P2P) model to integrate multiple data centers and their associated edge networks. The data center as cloud platform is responsible for the storage and processing of original RS images, as well as the storage of partial processed images while the edge network is mainly for caching and sharing the processed images. With this design, RS-pCloud not only achieves the load sharing of processing works but also attains the data efficiency among the edges at the same time, which in turn improves the performance of the imageprocessing and reduce the cost of the data transmission as well. RS-pCloud is designed to be used in a transparent way where it receives a query task from the user through a certain cloud platform, split the task into different sub-tasks, according to the location of the data they required, and then distribute the sub-tasks to corresponding clouds for near-data processing, the returned results from each cloud are first cached in specific edge for further sharing and then gathered at the client to synthesize the final product. We implemented and deployed RS-pCloud on three clusters in conjunction with an edge network to show its performance advantages over traditional single-cluster systems.
Cross-modal matching, which aims to establish the correspondence between two different modalities, is fundamental to a variety of tasks such as cross-modal retrieval and vision-and-language understanding. Although a h...
ISBN:
(纸本)9781713845393
Cross-modal matching, which aims to establish the correspondence between two different modalities, is fundamental to a variety of tasks such as cross-modal retrieval and vision-and-language understanding. Although a huge number of cross-modal matching methods have been proposed and achieved remarkable progress in recent years, almost all of these methods implicitly assume that the multimodal training data are correctly aligned. In practice, however, such an assumption is extremely expensive even impossible to satisfy. Based on this observation, we reveal and study a latent and challenging direction in cross-modal matching, named noisy correspondence, which could be regarded as a new paradigm of noisy labels. Different from the traditional noisy labels which mainly refer to the errors in category labels, our noisy correspondence refers to the mismatch paired samples. To solve this new problem, we propose a novel method for learning with noisy correspondence, named Noisy Correspondence Rectifier (NCR). In brief, NCR divides the data into clean and noisy partitions based on the memorization effect of neural networks and then rectifies the correspondence via an adaptive prediction model in a co-teaching manner. To verify the effectiveness of our method, we conduct experiments by using the image-text matching as a showcase. Extensive experiments on Flickr30K, MS-COCO, and Conceptual Captions verify the effectiveness of our method.
The fuzzy matrix probabilistic neural network (FMPNN) designed to solving image classification tasks where images are represented in matrix form is proposed. This neuro-fuzzy system has four layers of data processing ...
详细信息
ISBN:
(数字)9781728132143
ISBN:
(纸本)9781728132150
The fuzzy matrix probabilistic neural network (FMPNN) designed to solving image classification tasks where images are represented in matrix form is proposed. This neuro-fuzzy system has four layers of data processing which are fed into the system in data stream form on the "sliding window". It is supposed that formed classes of data-images can be arbitrarily overlapped in feature space creating a situation of fuzziness. The tuning of FMPNN is the combination of learning procedures such as "Lazy learning", "Winner takes all", "Learning vector quantization", "Fuzzy C-means clustering". The feature of the proposed fuzzy classification - image recognition system is the high speed of data processing allowing to process data within the concept of Data Stream Mining.
Muon imaging has been used more and more in recent years. Based on the principle of multiple Coulomb scattering and the Molier formula, by detecting the direction of the incident muon and the exit muon, and according ...
详细信息
ISBN:
(数字)9781665421133
ISBN:
(纸本)9781665421140
Muon imaging has been used more and more in recent years. Based on the principle of multiple Coulomb scattering and the Molier formula, by detecting the direction of the incident muon and the exit muon, and according to a specific muon imaging algorithm, cosmic ray muon imaging can realize high-Z material identification and imaging. The high position resolution cosmic ray muon detection system can detect the incident and exiting positions of the muon more accurately, and can realize the material identification in a more detailed manner. Different detector designs will affect the position resolution of the detector systems and then affect the final imaging results. In this paper, different detectors are designed and simulated, and a high-resolution cosmic ray imaging system is obtained. In addition, different algorithm processing will also affect the imaging quality. The track reconstruction algorithm PoCA(The point of closest approach) based on scattering angle is used to reconstruct the system simulation results. And we put forward the direction of algorithm optimization.
image recognition has always been an important research topic in the field of pattern recognition. This article uses the printed character recognition of bills as the research background. Through the detailed analysis...
详细信息
image recognition has always been an important research topic in the field of pattern recognition. This article uses the printed character recognition of bills as the research background. Through the detailed analysis of the recognition process, some preprocessingalgorithms are selected and implemented, including image enhancement, binarization, filtering, character cutting, skew correction, etc. Finally, transplant each algorithm to the simulation platform, mainly for BP algorithm, according to the characteristics of DSP, optimize from two aspects of compilation options and source code to meet the requirements of fast processing. Finally, the requirements of better recognition effect and fast processing speed are realized.
Security has become an increasingly popular domain in computer science. Because today devices are attacked not only by classical viruses but also intentional attacks occur, in order to stole data from devices and obta...
详细信息
ISBN:
(纸本)9781728107738
Security has become an increasingly popular domain in computer science. Because today devices are attacked not only by classical viruses but also intentional attacks occur, in order to stole data from devices and obtain material benefits, it is necessary to add new security types. These new ways of securing applications include technologies such as facial recognition, iris recoil, fingerprint recognition. Fingerprint algorithms have been embedded in modern mobile devices, for authentication. Nowadays mobile devices use a fingerprint scanner, so fingerprinting algorithms are becoming more widespread. This paper presents a fingerprint authentication algorithm based on the Advance Encryption Standard (AES) and on the Android Keystore System. The proposed algorithm is tested, by integrating it into a budget application.
The authors studied the task of processing the information from the optical system when the UAV is landing on a moving unmanned vehicle. Generally, color image analysis algorithms are very accurate, but they cannot wo...
详细信息
ISBN:
(数字)9785919950684
ISBN:
(纸本)9781728187983
The authors studied the task of processing the information from the optical system when the UAV is landing on a moving unmanned vehicle. Generally, color image analysis algorithms are very accurate, but they cannot work in real time or need to enhance the performance of professional computers. A compact high-speed color image recognition algorithm is developed basing on a pre-processing method-a "downsample" function for decimation; HSV model; Otsu's method - an algorithm for calculating the binary threshold of grayscale images, and method for isolating connected components-Two-Pass method. The simulation results demonstrated the operating capability and high enough efficiency of the developed algorithm. It is possible to achieve a significant reduction in the implementation time of the algorithm by using the decimation function and the HSV model.
Edge detection is one of the significant properties of many computer vision systems in image recognition, enhancement, compression, restoration, and so on. Though many research results were published, researchers stil...
详细信息
ISBN:
(数字)9781728193335
ISBN:
(纸本)9781728193342
Edge detection is one of the significant properties of many computer vision systems in image recognition, enhancement, compression, restoration, and so on. Though many research results were published, researchers still try to develop a more efficient edge detection algorithm that meets the current needs. We propose an image pre-processing based edge detection technique using canny edge detector. The main problem of canny edge detector is that it cannot identify the edges which are slightly vague due to the Gaussian smoothing. In this study, we solve the problem using the histogram processing of an image before using the canny edge detector. The result demonstrates that the proposed approach gives better outcomes compared to the state-of-the-art edge detectors.
暂无评论