In online insurance, one of the central challenges is the cold-starting of new insurance products, which means there are no previous samples to refer to. Previous studies have mainly focused on improving the predictio...
Protein secondary structure prediction is still a challenging task in bioinformatics, especially for 8-state (Q8) classification. To address this problem, we have proposed a deep learning based model by integrating gr...
详细信息
Underwater optical imaging produces images with high resolution and abundant information and hence has outstanding advantages in short-distance underwater target ***,low-light and high-noise scenarios pose great chall...
详细信息
Underwater optical imaging produces images with high resolution and abundant information and hence has outstanding advantages in short-distance underwater target ***,low-light and high-noise scenarios pose great challenges in un-derwater image and video *** improve the accuracy and anti-noise performance of underwater target image edge detection,an underwater target edge detection method based on ant colony optimization and reinforcement learning is proposed in this ***,the reinforcement learning concept is integrated into artificial ants’movements,and a variable radius sensing strategy is pro-posed to calculate the transition probability of each *** methods aim to avoid undetection and misdetection of some pixels in image ***,a double-population ant colony strategy is proposed,where the search process takes into account global search and local search *** results show that the algorithm can effectively extract the contour information of underwater targets and keep the image texture well and also has ideal anti-interference performance.
Semi-supervised multi-label learning (SSMLL) refers to inducing classifiers using a small number of samples with multiple labels and many unlabeled samples. The prevalent solution of SSMLL involves forming pseudo-labe...
Short text classification has gained significant attention in the information age due to its prevalence and real-world applications. Recent advancements in graph learning combined with contrastive learning have shown ...
详细信息
3D convolutions are commonly employed by demosaicking neural models, in the same way as solving other image restoration problems. Counter-intuitively, we show that 3D convolutions implicitly impede the RGB color spect...
详细信息
Automatic heart sound diagnosis plays an important role in the early detection of cardiovascular diseases. Phonocardiogram (PCG) signals are often used in this field f or its low cost and non-invasive advantages. In t...
详细信息
Despite the great success of spoken language understanding (SLU) in high-resource languages, it remains challenging in low-resource languages mainly due to the lack of labeled training data. The recent multilingual co...
详细信息
In this paper, we take Mongolian ethnic patterns as the research object and conduct a technical study on ethnic pattern generation and super-resolution reconstruction based on deep learning generative adversarial netw...
In this paper, we take Mongolian ethnic patterns as the research object and conduct a technical study on ethnic pattern generation and super-resolution reconstruction based on deep learning generative adversarial networks to address the problems of low quality of ethnic pattern data collection and lack of innovation. In this paper, we use a database including 1621 Mongolian ethnic patterns, train StyleGAN2 on the pre-processed images, and generate images After that, we use ESRGAN to perform super-resolution reconstruction on the generated images to generate high-resolution patterns with Mongolian style. The model designed in this paper can generate high quality images with ethnic characteristics in the case of insufficient original data. Compared with the time-consuming and labor-intensive traditional ethnic pattern design methods, the model designed in this paper lowers the threshold of ethnic pattern innovation and contributes to the innovative design of ethnic patterns to a certain extent, which has some positive significance for the protection and inheritance of ethnic patterns.
In the heavy clutter environment, the information capacity is large,the relationships among information are complicated, and track initiationoften has a high false alarm rate or missing alarm rate. Obviously, it is ad...
详细信息
In the heavy clutter environment, the information capacity is large,the relationships among information are complicated, and track initiationoften has a high false alarm rate or missing alarm rate. Obviously, it is adifficult task to get a high-quality track initiation in the limited measurementcycles. This paper studies the multi-target track initiation in heavy *** first, a relaxed logic-based clutter filter algorithm is presented. In thealgorithm, the raw measurement is filtered by using the relaxed logic *** not only design a kind of incremental and adaptive filtering gate, but alsoadd the angle extrapolation based on polynomial extrapolation. The algorithm eliminates most of the clutter and obtains the environment with highdetection rate and less clutter. Then, we propose a fuzzy sequential Houghtransform-based track initiation algorithm. The algorithm establishes a newmeshing rule according to system noise to balance the relationship between thegrid granularity and the track initiation quality. And a flexible superpositionmatrix based on fuzzy clustering is constructed, which avoids the transformation error caused by 0–1 voting method in traditional Hough *** addition, the algorithm allows the superposition matrixes of nonadjacentcycles to be associated to overcome the shortcoming that the track can’t beinitiated in time when the measurements appear in an intermittent way. Anda slope verification method is introduced to detect formation-intensive serialtracks. Last, the sliding window method is employed to feedback the trackinitiation results timely and confirm the track. Simulation results verify thatthe proposed algorithms can initiate the tracks accurately in heavy clutter.
暂无评论