Search

Rectified activation units make an important contribution to the success of deep neural networks in many computer vision tasks. In this paper, we propose a Parametric Deformable Exponential Linear Unit (PDELU) and theoretically verify its effectiveness for improving the convergence speed of learning procedure. By means of flexible map shape, the proposed PDELU could push the mean value of activation responses closer to zero, which ensures the steepest descent in training a deep neural network. We verify the effectiveness of the proposed method in the image classification task. Extensive experiments on three classical databases (i.e., CIFAR-10, CIFAR-100, and ImageNet-2015) indicate that the proposed method leads to higher convergence speed and better accuracy when it is embedded into different CNN architectures (i.e., NIN, ResNet, WRN, and DenseNet). Meanwhile, the proposed PDELU outperforms many existing shape-specific activation functions (i.e., Maxout, ReLU, LeakyReLU, ELU, SELU, SoftPlus, Swish) and the shape-adaptive activation functions (i.e., APL, PReLU, MPELU, FReLU). [ABSTRACT FROM AUTHOR]

In this paper, a novel reduced reference (RR) stereoscopic image quality assessment (SIQA) is proposed by characterizing the statistical properties of the stereoscopic image in the reorganized discrete cosine transform (RDCT) domain. Firstly, the difference image between the left and right view images is computed. Afterwards, the left and right view images, as well as the difference image, are decomposed by block-based discrete cosine transform (DCT). The DCT coefficients are further reorganized into a three-level coefficient tree, resulting in ten RDCT subbands. For each RDCT subband, the statistical property of the coefficient distribution is modeled by the generalized Gaussian density (GGD) function. And the mutual information (MI) and energy distribution ratio (EDR) are employed to depict the statistical properties across different RDCT subbands. Moreover, EDR can further model the mutual masking property of the human visual system (HVS). By considering the GGD modeling behavior within each RDCT subband and MI together EDR characterizing behavior across RDCT subbands, the statistical properties of the stereoscopic image are fully exploited, including the left view, right view, and the difference image. Experimental results demonstrated that the statistical properties of the difference image can well represent the perceptual quality of the stereoscopic image, which outperforms the representative RR quality metrics for stereoscopic image and even some full reference (FR) quality metrics. By considering the left view, right view, and difference image together, the performances of the proposed RR SIQA can be further improved, which presenting a more closely relationship between the quality metric output and human visual perception. [ABSTRACT FROM AUTHOR]

In this paper, we present a hybrid human recognition system for surveillance. A Cascade Head–Shoulder Detector (CHSD) with human body model is proposed to find the face region in a surveillance video frame image. The CHSD is a chain of rejecters which combines the advantages of Haar-like feature and HoG feature to make the detector more efficient and effective. For human recognition, we introduce an Overlapping Local Phase Feature (OLPF) to describe the face region, which can improve the robustness to pose change and blurring. To well model the variations of faces, an Adaptive Gaussian Mixture Model (AGMM) is presented to describe the distributions of the face images. Since AGMM does not need the facial topology, the proposed method is resistant to face detection error caused by imperfect localization or misalignment. Experimental results demonstrate the effectiveness of the proposed method in public dataset as well as real surveillance video. [ABSTRACT FROM AUTHOR]