Diffusion-based personalized visual content generation technologies have achieved significant breakthroughs, allowing for the creation of specific objects by just learning from a few reference photos. However, when misused to fabricate fake news or unsettling content targeting individuals, these technologies could cause considerable societal harm. To address this problem, current methods generate adversarial samples by adversarially maximizing the training loss, thereby disrupting the output of any personalized generation model trained with these samples. However, the existing methods fail to achieve effective defense and maintain stealthiness, as they overlook the intrinsic properties of diffusion models. In this paper, we introduce a novel Dual-Domain Anti-Personalization framework (DDAP). Specifically, we have developed Spatial Perturbation Learning (SPL) by exploiting the fixed and perturbation-sensitive nature of the image encoder in personalized generation. Subsequently, we have designed a Frequency Perturbation Learning (FPL) method that utilizes the characteristics of diffusion models in the frequency domain. The SPL disrupts the overall texture of the generated images, while the FPL focuses on image details. By alternating between these two methods, we construct the DDAP framework, effectively harnessing the strengths of both domains. To further enhance the visual quality of the adversarial samples, we design a localization module to accurately capture attentive areas while ensuring the effectiveness of the attack and avoiding unnecessary disturbances in the background. Extensive experiments on facial benchmarks have shown that the proposed DDAP enhances the disruption of personalized generation models while also maintaining high quality in adversarial samples, making it more effective in protecting privacy in practical applications.

Recognizing the actions performed in a video is challenging for an intelligent system since there are wide variations and enormous information in the video. Attention mechanism pays attention to key target areas, ignores irrelevant information and extracts more discriminant features. In recent years, attention mechanism has been introduced into video recognition. Although a rich literature has been spawned, most of the research on attention aims to aggregate local features by attention. Instead of feature aggregation, we propose to aggregate decisions based on local spatio-temporal attention regions for action recognition, which is inspired by ensemble learning. The proposed decision fusion module is easy to interpret and architecture-independent. In this article, the regions around the body joints are regarded as the key regions. We use the corresponding regions of the body joints in the 3-D feature maps as the basic local features for local classification. Finally, all the local classification results are combined to make a global decision. Furthermore, when training the network, we can selectively add supervision to the local and global decisions. We experimentally show that the proposed mechanism can improve the recognition performance on multiple datasets which demonstrates its effectiveness.

Large-scale 3D Point cloud classification is a basic topic for various applications. Traditional geometries features are usually independent of each other and difficult to adapt to a fixed classification model. With the rise of the neural network, deep learning is considered in 3D point cloud application. 3D points are difficult to feed the neural network directly based on deep learning, as they cannot be arranged in a fixed order as image pixels. In this paper, we combine traditional feature-based methods with the Convolutional neural network(CNN) to finish the classification task. The core idea is to construct a feasible structure called Feature Description Matrix(FDM) which encapsulates the local feature of the point to feed CNN for training and testing. By extracting geometry features and designed Feature Description Vectors(FDV) for FDM, a simple mechanism for point cloud classification is given, and experiments validate the effectiveness of our method, with higher classification accuracy compared to state-of-art works.

In this paper we present a novel silhouette-based method to estimate 2D human pose. It takes a pre-defined human skeleton model as the prior information and a video sequence as the data source, and estimates human pose in each frame by the following steps: Firstly, the Gaussian Mixture Background Model (GMM) is adopted to extract silhouette from an image and this silhouette will be the human body data set after treatment. Then the Distance Transform (DT) and Principal Component Analysis (PCA) are introduced to locate the base point of the human skeleton model, with which the human skeleton model is initialized automatically. Afterwards an iterative process, based on the Expectation Maximization (EM), is constructed out to cluster the human body data set and estimate the parameters of the skeleton iteratively. Finally the pose of human body is figured out in the form of a corresponding skeleton model. Extensive experiments show that this method is robust and precise, and is feasible to apply in the real-time system.

Multi-view approach has been proposed to solve occlusion and lack of visibility in crowded scenes. However, the problem is that too much redundancy information might bring about false alarm. Although researchers have done many efforts on how to use the multi-view information to track people accurately, it is particularly hard to wipe off the false alarm. Our approach is to use multiple views cooperatively to detect objects and use objects silhouette on planes of different height to remove false alarm. To achieve this we adopt a novel multi-planar homography constraint to resolve occlusions and false alarm. Experimental results show that our algorithm is able to accurately locate people in crowded scene maintaining correct correspondences across views. Moreover, the false alarm rate is obviously reduced.

To against the complexity in finding feature-pair in image registration caused by traditional features: corners, lines and image edge, a novel method for image registration based on rectangle pattern is proposed. Unlike traditional features, a rectangle pattern can be described as its four vertexes and center, which can afford five pair-wise points for any kind of image transformation, and it holds stable in different weather condition, time and imaging way, which formed by building angular, widely found in aero-image. Firstly, image edge is detected by canny, and then distance transform is applied on the result of canny, by follows, a thresholding and mask convoluting are used on the distance transform result to avoid the interfering complex lines and edge. The center of the rectangles is obtained by clustering on the prior result, and then the four vertexes is calculated by geometric restrict on rectangle pattern. Consequently, we calculate the centers pair of the rectangles by their slope difference which aims to get the correct pair-wise points set between two images. Finally, we use the four vertexes pair and center pair of the rectangle pattern as the input of RANSAC algorithm, to solve an affine transform. The proposed algorithm is proved to be effective and accurate on the translation, rotation and scale between electro-optic(EO) images pair and SAR-EO images pair. 1.3 pixels of registration accuracy result is obtained in the experiment.