Search

Observing that the existing model compression approaches only focus on reducing the redundancies in convolutional neural networks (CNNs) along one particular dimension (e.g., the channel or spatial or temporal dimension), in this work, we propose our multidimensional pruning (MDP) framework, which can compress both 2-D CNNs and 3-D CNNs along multiple dimensions in an end-to-end fashion. Specifically, MDP indicates the simultaneous reduction of channels and more redundancy on other additional dimensions. The redundancy of additional dimensions depends on the input data, i.e., spatial dimension for 2-D CNNs when using images as the input data, and spatial and temporal dimensions for 3-D CNNs when using videos as the input data. We further extend our MDP framework to the MDP-Point approach for compressing point cloud neural networks (PCNNs) whose inputs are irregular point clouds (e.g., PointNet). In this case, the redundancy along the additional dimension indicates the point dimension (i.e., the number of points). Comprehensive experiments on six benchmark datasets demonstrate the effectiveness of our MDP framework and its extended version MDP-Point for compressing CNNs and PCNNs, respectively.

Unsupervised domain adaption (UDA) aims to reduce the domain gap between labeled source and unlabeled target domains. Many prior works exploit adversarial learning that leverages pre-designed discriminators to drive the network for aligning distributions between domains. However, most of them do not consider the degeneration of the domain discriminators caused by the gradually dominating gradients of aligned target samples during training, and they still suffer from the cross-domain semantic mismatch problem in the learned feature space. Hence, this paper attempts to understand and solve both issues from the lens of optimization loss and propose an automatic loss function search for adversarial domain adaptation (ALSDA). First, we extend the common adversarial loss by adding an adjustable hyper-parameter that can re-weight the gradients assigned to target samples, so that the domain discriminator can impose consecutive and influential driving forces for domain alignment. Meanwhile, we upgrade the traditional orthogonality loss with class-wisely adjustable hyper-parameters that can strengthen the cross-domain feature separation. Since manually determining the optimal loss functions requires expensive expert efforts, we leverage the popular AutoML to automatically search for the optimal loss functions from a pre-defined novel and unique search space for UDA. Further, to enable the loss function search when the target domain is unlabeled, we introduce a simple-but-effective entropy-guided search strategy with the aid of REINFORCE learning. Extensive experiments on various typical baselines and benchmark datasets such as Office-Home, Office-31, and Birds-31 have been conducted, and the results validate the generalization and superiority of the proposed ALSDA.

The existing end-to-end optimized 3D action recognition methods often suffer from high computational costs. Observing that different frames and different points in point cloud sequences often have different importance values for the 3D action recognition task, in this work, we propose a fully automatic model compression framework called 3D-Pruning (3DP) for efficient 3D action recognition. After performing model compression by using our 3DP framework, the compressed model can process different frames and different points in each frame by using different computational complexities based on their importance values, in which both the importance value and computational complexity for each frame/point can be automatically learned. Extensive experiments on five benchmark datasets demonstrate the effectiveness of our 3DP framework for model compression.

This review covers the decomposition mechanisms of various printing ink binder resins, with a particular focus on their behavior under extrusion conditions in the mechanical recycling process of polyolefin (PO) based plastic packaging. Thermal degradation and hydrolysis of the nitrocellulose (NC) — the most used binder for flexographic surface printing on single-layer flexible plastic packaging, occur concurrently during the mechanical recycling process under 160–210 °C. For other printing ink binders, polyurethane (PU) noticeable degradation takes place between 200 and 300 °C, mostly above 250 °C. However, with the involvement of humidity, degradation by hydrolysis can start from 150 °C. A similar effect is also discovered with the cellulose acetate (CA) derivatives, which are thermally stable until 300 °C and can be hydrolyzed at 100 °C. The thermal stability of polyvinyl butyral (PVB) is not influenced by humidity, with thermal stability ranging from 170 to 260 °C, depending on different types. Ultraviolet (UV)-cured acrylics are thermally stable until 400 °C. The hydrolysis degradation can take place at room temperature. Moreover, this review covers the thermal stability of different colorants used for printing ink application and elaborates on several thermal-stable alternatives of some common colors. This study further reviews how the binder resin affects the quality of recyclates, revealing it to be not only induced by the degradation of the binder resin but also by the immiscibility between the plastic and binder resin. In advanced recycling processes, mainly selective dissolution-precipitation and pyrolysis, the presence of binder resin and its degradation products could still affect the quality of the product. This review accentuates the imperative need for in-depth research to unravel the impact of printing ink constituents on the quality of recycled products.

Cloud vendors are actively adopting FPGAs into their infrastructures for enhancing performance and efficiency. As cloud services continue to evolve, FPGA (field programmable gate array) systems would play an even important role in the future. In this context, FPGA sharing in multi-tenancy scenarios is crucial for the wide adoption of FPGA in the cloud. Recently, many works have been done towards effective FPGA sharing at different layers of the cloud computing stack.

Detecting following events is important for intelligent video surveillance systems, especially at secure entrances. However, there is no effective approach for recognizing following actions. In this paper, we propose a novel following action recognition method based on motion analysis of the human skeleton. We use a Kinect to capture the human skeleton and represent the following action as a joint-trajectory difference feature. The most difficult aspect of recognizing a following action lies in classifying both the following action and the companion action. To avoid false alarms from companion actions, this type of human interaction should be detected by the relative displacement of joint points. In our approach, an SVM classifier is trained to recognize following actions, companion actions and others. To verify the proposed method, we built a new following action dataset during this study. The experimental results show that the proposed method is effective at recognizing following actions.