Image matting is a fundamental technique used to extract a fine foreground image from a given image by estimating the opacity values of each pixel. It is one of the key techniques in image processing and has a wide range of applications in practical scenarios, such as in image and video editing. Deep learning has demonstrated outstanding performance in various image processing tasks, making it a popular research topic. In recent years, image matting methods based on deep learning have gained significant attention due to their superior performance. Therefore, this article presents a comprehensive overview of the deep learning-based image matting algorithms that have been proposed in recent years. This paper initially introduces frequently used datasets and their production methods, along with the basic principles of traditional image matting techniques. We then analyze deep learning-based matting algorithms in detail and introduce commonly used image matting evaluation metrics. Additionally, this paper discusses the application scenarios of image matting, conducts experiments to illustrate the limitations of current image matting methods, and outlines potential future research directions in this field. Overall, this paper can serve as a valuable reference for researchers that are interested in image matting.

In wearable human activity recognition (HAR), the generation and utilization of virtual IMU data has recently gained attention. The use of virtual data can improve the robustness, effective features, and customized motion types chosen in the HAR system. However, few studies have focused on augmenting virtual IMU data to reduce the dependence on real IMU data during the machine learning model training phase. This work proposes modifying the reconstructed 3D motion by its joint data to generate larger 3D motion based on a given 3D motion sequence, thus augmenting the virtual IMU dataset. The method simulates the intra-difference of the same motion type in the real world. It aims to use fewer 3D motion inputs to generate the larger size of the virtual IMU dataset to recognize the real activity. The experiment demonstrated the feasibility of the proposed method and provided insight into the 3D motion modification-based augmentation method.

Building a wearable human activity recognition (HAR) system follows complicated steps, making the wearable HAR system development time-consuming, cumbersome, and error-prone. One of the critical challenges is its dataset collection, which typically requires a group of people to perform specified motions for an extensive period to capture data. Recently, cross-modal virtual IMU data generation from 3D avatars has been recognized as a key solution in addressing HAR system establishment’s costly and laborious sensor dataset collection. However, the utilization of virtual IMU data still faces the challenge of tedious steps, including expensive computing resources, model training, and domain transfer, among others. This paper presents a novel platform employing the input of 3D avatar motion sequences. Combined with the virtual IMU data generation mechanism, the motion modification-based virtual IMU data augmentation approach is designed to generate the virtual IMU training dataset. The platform supports various sensor placement selection, data processing, model training, fine-tuning, and deployment functionalities. The proposed platform integrates the whole process of developing a wearable HAR system. Due to the fewer requirements in the real world, it could enable the wearable HAR system building to become low-cost, customized and contribute to HAR system prototyping and related application scenarios.