Face Manipulation Detection Based on Supervised Multi-Feature Fusion Attention Network

Nowadays, faces in videos can be easily replaced with the development of deep learning, and these manipulated videos are realistic and cannot be distinguished by human eyes. Some people maliciously use the technology to attack others, especially celebrities and politicians, causing destructive social impacts. Therefore, it is imperative to design an accurate method for detecting face manipulation. However, most of the existing methods adopt single convolutional neural network as the feature extraction module, causing the extracted features to be inconsistent with the human visual mechanism. Moreover, the rich details and semantic information cannot be reflected with single feature, limiting the detection performance. Therefore, this paper tackles the above problems by proposing a novel face manipulation detection method based on a supervised multi-feature fusion attention network (SMFAN). Specifically, the capsule network is used for face manipulation detection, and the SMFAN is added to the original capsule network to extract details of the fake face image. Further, the focal loss is used to realize hard example mining. Finally, the experimental results on the public dataset FaceForensics++ show that the proposed method has better performance.

The usage of deep learning algorithm in medical diagnostic of breast cancer

Diagnosis is a crucial step to identify the disease that experienced by the patient. Diagnosis includes information gathering, integration, and interpretation. However, diagnosis process is not an easy task. Diagnostic accuracy is depending on the experience and cognitive ability of diagnosticians. The new algorithm called deep learning that is developed by simulating the human visual mechanism has been implemented in medical diagnostics. One of the diseases that can be diagnosed by using deep learning algorithm is the breast cancer. Several studies showed that deep learning algorithm can be used for detecting and classifying lesions, detecting mitosis, and predicting specific gene status. In this review article, 16 research journals were reviewed and discussed. The limitations of each algorithm are provided. All of the journals showed that deep learning algorithm has high diagnostics accuracy in assisting the professional diagnosticians to determine diagnosis outcome accordingly.

Research on Shape Feature Recognition of Boundary Representation CAD Model

B-Rep (Boundary Representation) CAD model is being widely used in representation of industrial product, so its feature recognition has acquired widespread research interests in computer vision and 3D model retrieval fields. We present one approach of feature recognition based on the idea of human visual mechanism. Surfaces as the visual shape features, and solids as well as shells as the topological relations, were extracted from the neutral STEP (Standard for Exchange of Product Model Data) files of B-Rep model. Towards three surface types of NURBS, analytical and poly loop, the properties of surface boundary and region are established based on curvature and other geometric index. So B-Rep CAD model is characterized as the hierarchical tree with solid layer, shell layer and surface layer for object recognition and retrieval, and the corresponding experiments verified the effectiveness of the method of shape feature recognition.

Information Processing in Visual System - Optical Illusion and Visual Mechanism

Human beings obtain big amount of information from the external world through their visual system. Automated system such as robot must provide the visual functions for their flexible operations in 3-D circumstances. In order to realize the visual function artificially, we would be better to learn from the human visual mechanism. Optical illusions would be a pure reflection of the human visual mechanism; they can be used for investigating human visual mechanism. New types of optical illusion with binocular viewing are introduced and investigated.

A Study on Visual Mechanism with Optical Illusions

Optical illusion seems to be the phenomena which are purely reflecting the mechanism of.human visual system and are expected as the effective cues to elucidating human visual mechanism. The author found the new types of 3-D visual illusion with binocular viewing. From the visual stimuli of binocular disparity given only along the contour of an object, human visual system can perceive entire 3-D illusory object where there are no physical visual stimuli giving depth information. They have close relation with the 3-D space perceiving functions in the human visual system. A study on these newly found optical illusions are introduced and the considerations are made for their applications and the exploitations including the contributions of information processing techniques such as computer graphics, computer vision and so on.

3-D Computer Graphics System for Vision Research by Binocular Viewing

For the artificial realization and the application of the visual function of 3-D space perception, the better understanding of human visual mechanism is required strongly. The disparity and occlusion observed with binocular viewing seems to be the most important cues to get 3-D information. Then, the authors developed a simple stereoscopic display system using a time sharing display of left eye and right eye images with liquid crystal shutter. This system is composed of a simple small control circuit and has big advantages such as that the hardware can be applied to any types of display system and the software can be transplant to different type computer system easily. Then, the authors tried to apply this system for the vision research of 3-D perceptural function with binocular viewing.