arXiv subjects

Yanhui Guo

Publications and source records attributed to Yanhui Guo.

22 records · Page 2Linked to original sources

arXiv2020

Deep Multi-modality Soft-decoding of Very Low Bit-rate Face Videos

We propose a novel deep multi-modality neural network for restoring very low bit rate videos of talking heads. Such video contents are very common in social media, teleconferencing, distance education, tele-medicine, etc., and often need to be transmitted with limited bandwidth. The proposed CNN method exploits the correlations among three modalities, video, audio and emotion state of the speaker, to remove the video compression artifacts caused by spatial down sampling and quantization. The deep learning approach turns out to be ideally suited for the video restoration task, as the complex non-linear cross-modality correlations are very difficult to model analytically and explicitly. The new method is a video post processor that can significantly boost the perceptual quality of aggressively compressed talking head videos, while being fully compatible with all existing video compression standards.

cs.CV↗

arXiv2018

Certainty of outlier and boundary points processing in data mining

Data certainty is one of the issues in the real-world applications which is caused by unwanted noise in data. Recently, more attentions have been paid to overcome this problem. We proposed a new method based on neutrosophic set (NS) theory to detect boundary and outlier points as challenging points in clustering methods. Generally, firstly, a certainty value is assigned to data points based on the proposed definition in NS. Then, certainty set is presented for the proposed cost function in NS domain by considering a set of main clusters and noise cluster. After that, the proposed cost function is minimized by gradient descent method. Data points are clustered based on their membership degrees. Outlier points are assigned to noise cluster and boundary points are assigned to main clusters with almost same membership degrees. To show the effectiveness of the proposed method, two types of datasets including 3 datasets in Scatter type and 4 datasets in UCI type are used. Results demonstrate that the proposed cost function handles boundary and outlier points with more accurate membership degrees and outperforms existing state of the art clustering methods.

eess.SP↗

arXiv2018

An effective clustering method based on data indeterminacy in neutrosophic set domain

In this work, a new clustering algorithm is proposed based on neutrosophic set (NS) theory. The main contribution is to use NS to handle boundary and outlier points as challenging points of clustering methods. In the first step, a new definition of data indeterminacy (indeterminacy set) is proposed in NS domain based on density properties of data. Lower indeterminacy is assigned to data points in dense regions and vice versa. In the second step, indeterminacy set is presented for a proposed cost function in NS domain by considering a set of main clusters and a noisy cluster. In the proposed cost function, two conditions based on distance from cluster centers and value of indeterminacy, are considered for each data point. In the third step, the proposed cost function is minimized by gradient descend methods. Data points are clustered based on their membership degrees. Outlier points are assigned to noise cluster; and boundary points are assigned to main clusters with almost same membership degrees. To show the effectiveness of the proposed method, three types of datasets including diamond, UCI and image datasets are used. Results demonstrate that the proposed cost function handles boundary and outlier points with more accurate membership degrees and outperforms existing state of the art clustering methods in all datasets.

eess.IV↗

arXiv2018

Multiple Convolutional Neural Network for Skin Dermoscopic Image Classification

Melanoma classification is a serious stage to identify the skin disease. It is considered a challenging process due to the intra-class discrepancy of melanomas, skin lesions low contrast, and the artifacts in the dermoscopy images, including noise, existence of hair, air bubbles, and the similarity between melanoma and non-melanoma cases. To solve these problems, we propose a novel multiple convolution neural network model (MCNN) to classify different seven disease types in dermoscopic images, where several models were trained separately using an additive sample learning strategy. The MCNN model is trained and tested using the training and validation sets from the International Skin Imaging Collaboration (ISIC 2018), respectively. The receiver operating characteristic (ROC) curve is used to evaluate the performance of the proposed method. The values of AUC (the area under the ROC curve) were used to evaluate the performance of the MCNN.

cs.CV↗