Double-Shift: A Low-Power DNN Weights Storage and Access Framework based on Approximate Decomposition and Quantization

One major challenge in deploying Deep Neural Network (DNN) in resource-constrained applications, such as edge nodes, mobile embedded systems, and IoT devices, is its high energy cost. The emerging approximate computing methodology can effectively reduce the energy consumption during the computing process in DNN. However, a recent study shows that the weight storage and access operations can dominate DNN's energy consumption due to the fact that the huge size of DNN weights must be stored in the high-energy-cost DRAM. In this paper, we propose Double-Shift, a low-power DNN weight storage and access framework, to solve this problem. Enabled by approximate decomposition and quantization, Double-Shift can reduce the data size of the weights effectively. By designing a novel weight storage allocation strategy, Double-Shift can boost the energy efficiency by trading the energy consuming weight storage and access operations for low-energy-cost computations. Our experimental results show that Double-Shift can reduce DNN weights to 3.96%–6.38% of the original size and achieve an energy saving of 86.47%–93.62%, while introducing a DNN classification error within 2%.

Fast Segmentation of Vertebrae CT Image Based on the SNIC Algorithm

Automatic image segmentation plays an important role in the fields of medical image processing so that these fields constantly put forward higher requirements for the accuracy and speed of segmentation. In order to improve the speed and performance of the segmentation algorithm of medical images, we propose a medical image segmentation algorithm based on simple non-iterative clustering (SNIC). Firstly, obtain the feature map of the image by extracting the texture information of it with feature extraction algorithm; Secondly, reduce the image to a quarter of the original image size by downscaling; Then, the SNIC super-pixel algorithm with texture information and adaptive parameters which used to segment the downscaling image to obtain the superpixel mark map; Finally, restore the superpixel labeled image to the original size through the idea of the nearest neighbor algorithm. Experimental results show that the algorithm uses an improved superpixel segmentation method on downscaling images, which can increase the segmentation speed when segmenting medical images, while ensuring excellent segmentation accuracy.

Physical Evaluation of Transfersome that Contains Pandan Leaves Extract (Pandanus amaryllifolius R.)

Pandan leaves have been researched and have effectiveness in the treatment of burns. The process of healing burns takes a long time and cause a hard tissue because it loses its elasticity, making it difficult to penetrate. In this study, pandanus leaves were formulated into the nanovesicle carrier system, namely trasfersom. Transfersomes have the ability to deform, namely the ability to reduce the particle size 5-10 times from the original size when passing through the gaps between cells so that transfersom can increase the penetration of active substances. The three formulas used are based on the ratio of concentrations of trasfersome vesicles, namely phospholipids and span 80. Formula 1 is (90:10), Formula 2 (85:15) and Formula 3 (80:20). The best formula is determined based on transfersom characterization, including particle size and PDI (solidispersity index), zeta potential, entrapment efficiency, deformability, and TEM particle morphology. The results showed that Formula 3 (80:20) is the most stable formula with an average particle size of 730.1 ± 4.9 nm, PDI value <0.7, zeta potential - 9.94 ± 1.02 mV, efficiency absorption 80.23%, and the deformability value 6.225.

A Novel Attention-Based Network for Fast Salient Object Detection

In the current salient object detection network, the most popular method is using U-shape structure. However, the massive number of parameters leads to more consumption of computing and storage resources which are not feasible to deploy on the limited memory device. Some others shallow layer network will not maintain the same accuracy compared with U-shape structure and the deep network structure with more parameters will not converge to a global minimum loss with great speed. To overcome all of these disadvantages, we propose a new deep convolution network architecture with three contributions: (1) using smaller convolution neural networks (CNNs) to compress the model in our improved salient object features compression and reinforcement extraction module (ISFCREM) to reduce parameters of the model. (2) introducing channel attention mechanism to weigh different channels for improving the ability of feature representation. (3) applying a new optimizer to accumulate the long-term gradient information during training to adaptively tune the learning rate. The results demonstrate that the proposed method can compress the model to 1/3 of the original size nearly without losing the accuracy and converging faster and more smoothly on six widely used datasets of salient object detection compared with the others models. Our code is published in https://gitee.com/binzhangbinzhangbin/code-a-novel-attention-based-network-for-fast-salientobject-detection.git

Cost-Effective Watermarking Scheme for Authentication of Digital Fundus Images in Healthcare Data Management

This paper presents a cost-effective watermarking scheme for the authentication of healthcare data management. The digital fundus images are one particular class of medical images and it is widely used for screening mass population, identifying early symptoms of various diseases in healthcare. The mass volume of such data and its management requires an effective authentication scheme, while it is exchanged on an open network. The proposed scheme uses a watermarking technique to authenticate the digital fundus images. The watermark is generated concerning the portions of the original image using Singular value decomposition (SVD) and the remaining portions are used for embedding. The embedding process uses interleaving concepts across the red and blue planes of the original images to make the number of embedding as constant. The constant number of embedding is fixed for the original size of the given image to make the scheme as computationally cost-effective. The experiment showed the maximum capacity of the proposed scheme is 329960 bits for an image of size 565x584x3. It modifies 43% of the total number of embedded pixels against jittering attacks at an average. Comparative analysis showed that the proposed scheme uses only 1/3 of the original image size for embedding by retaining good imperceptibility of 54 dB. The net performance of the proposed scheme is found to be constant and it makes a scheme as cost-effective.

The Temple of Demeter Chthonia at Hermione

This paper presents the results of an architectural survey of the foundations of a Classical temple, presumably that of Demeter Chthonia, located inside the chief sanctuary of the ancient city of Hermione. It also studies ancient architectural members built into the walls of the Taxiarches Church situated on top of the temple foundations. By analysing these material remains and connecting them to the observations of 19th-century travellers to Kastri (Hermione), the paper draws conclusions about the original size and appearance of the Temple of Demeter Chthonia.

Inhalable Microparticles Embedding Calcium Phosphate Nanoparticles for Heart Targeting: The Formulation Experimental Design

Inhalation of Calcium Phosphate nanoparticles (CaPs) has recently unmasked the potential of this nanomedicine for a respiratory lung-to-heart drug delivery targeting the myocardial cells. In this work, we investigated the development of a novel highly respirable dry powder embedding crystalline CaPs. Mannitol was selected as water soluble matrix excipient for constructing respirable dry microparticles by spray drying technique. A Quality by Design approach was applied for understanding the effect of the feed composition and spraying feed rate on typical quality attributes of inhalation powders. The in vitro aerodynamic behaviour of powders was evaluated using a medium resistance device. The inner structure and morphology of generated microparticles were also studied. The 1:4 ratio of CaPs/mannitol led to the generation of hollow microparticles, with the best aerodynamic performance. After microparticle dissolution, the released nanoparticles kept their original size.

Replication of Carbonate Reservoir Pores at the Original Size Using 3D Printing

Three-dimensional (3D) printing is a powerful tool that enables visualization, replication, and experimentation with natural porous rocks. Over 100 years, natural rocks have been a focus of studies on how fluids such as hydrocarbons, greenhouse gases, and water flow through their porous systems. Scale and resolution are among the most challenging factors for current 3D printing methods when attempting to replicate the pore architecture of natural porous media. Most 3D printing techniques have resolution restraints during fabrication that makes feature reproduction at the 1:1 scale almost impossible. A new developing technology that uses two-photon lithography and ultraviolet (UV) light curable resin allows for nanometer features to be 3D printed. However, the main challenge of this 3D printing method is the small size of the resulting model (less than 20 mm in each direction). This technical note presents a detailed workflow on how to fabricate a carbonate rock replica at the micron scale. To test this workflow, a pore network was obtained from tomographic data of a reservoir rock core located in Mexico (1 mm in diameter and 2 mm in height) and was 3D printed at the original size. This replica was subjected to tomographic and scanning electron imaging to verify the accuracy of pore geometry. Incorporating lithographic printing into novel rock experiments that concern multiscale, multiphysics models of fluid flow and deformation open an unprecedented opportunity for more controlled prediction of reservoir fluid dynamics, carbon capture and storage, and continuum mechanics.

Energy and Storage Optimization In Precision Livestock Farming

This work introduces some of the most widely usedcompression algorithms, and their relevance to the field oflivestock farming, which has been historically characterizedfor requiring menial and inefficient labor, introducingenvironmental. And also for lacking the scale andautomation that cutting edge technologies can provide. Bydoing this we will explain how this opens the door tolocations untouched by technology, and the generaladvantages, and possibilities that integrating patternrecognition models bring to the table. In addition, we willexplain the ins and outs of these compression algorithms,and our reasoning behind our decision to choose analgorithm to implement in our pattern recognition model.To solve this problem, Seam Carving, Image Scaling andRun-Length encoding were used. With them we compressedthe images an average of 17.5% of their original size in atime complexity of O(L*N*M). This research shows howyou can create an efficient compression algorithm for usagein PLF.

Effect of the addition of selected silicon fillers on Si- PSA shrinkage

The concept of shrinkage phenomenon is widely described in the available literature. With respect to pressure-sensitive adhesives (PSA) in general, the definition of shrinkage is understood to be "less than its original size" and is closely related to the crosslinking process and the effect of the crosslinker on the test adhesive. Shrinkage alongside adhesive properties (adhesion, tackiness) and mechanical (cohesion) is one of the most important characteristics of a self-adhesive adhesive. It is very important in terms of production when receiving, for example, decorative banners or self-adhesive films where crosslinked adhesive and thus shrinkage can affect the surface of the adhesive material and create deformations. In the case of PSA, the acceptable adhesive pressure shrinkage must not exceed 0.5 %. Contraction is an important criterion for assessing the aging resistance of PSA materials. There are no studies on the shrinkage of silicone pressure-sensitive adhesives in literature, but many references to carbon-based adhesives have been reported.