Reduction In Storage Size Research Articles

Efficient Eye State Detection for Driver Fatigue Monitoring Using Optimized YOLOv7-Tiny

This study refines the YOLOv7-tiny model through structured pruning and architectural fine-tuning, specifically for real-time eye state detection. By focusing on enhancing the model’s efficiency, particularly in environments with limited computational resources, this research contributes significantly to advancing driver monitoring systems, where timely and accurate detection of eye states such as openness or closure can prevent accidents caused by drowsiness or inattention. Structured pruning was utilized to simplify the YOLOv7-tiny model, reducing complexity and storage requirements. Subsequent fine-tuning involved adjustments to the model’s width and depth to further enhance processing speed and efficiency. The experimental outcomes reveal a pronounced reduction in storage size, of approximately 97%, accompanied by a sixfold increase in frames per second (FPS). Despite these substantial modifications, the model sustains high levels of precision, recall, and mean average precision (mAP). These improvements indicate a significant enhancement in both the speed and efficiency of the model, rendering it highly suitable for real-time applications where computational resources are limited.

An auxiliary diagnostic tool for common fundus diseases based on fundus color photography and light-weight classification models.

To evaluate the performance of two lightweight neural network models in the diagnosis of common fundus diseases and make comparison to another two classical models. A total of 16,000 color fundus photography were collected, including 2000 each of glaucoma, diabetic retinopathy (DR), high myopia, central retinal vein occlusion (CRVO), age-related macular degeneration (AMD), optic neuropathy, and central serous chorioretinopathy (CSC), in addition to 2000 normal fundus. Fundus photography was obtained from patients or physical examiners who visited the Ophthalmology Department of Beijing Tongren Hospital, Capital Medical University. Each fundus photography has been diagnosed and labeled by two professional ophthalmologists. Two classical classification models (ResNet152 and DenseNet121), and two lightweight classification models (MobileNetV3 and ShufflenetV2), were trained. Area under the curve (AUC), sensitivity, specificity, accuracy, positive predictive value, and negative predictive value were used to evaluate the performance of the four models. Compared with the classical classification model, the total size and number of parameters of the two lightweight classification models were significantly reduced, and the classification speed was sharply improved. Compared with the DenseNet121 model, the ShufflenetV2 model took 50.7% less time to make a diagnosis on a fundus photography. The classical models performed better than lightweight classification models, and Densenet121 showed highest AUC in five out of the seven common fundus diseases. However, the performance of lightweight classification models is satisfying. The AUCs using MobileNetV3 model to diagnose AMD, diabetic retinopathy, glaucoma, CRVO, high myopia, optic atrophy, and CSC were 0.805, 0.892, 0.866, 0.812, 0.887, 0.868, and 0.803, respectively. For ShufflenetV2model, the AUCs for the above seven diseases were 0.856, 0.893, 0.855, 0.884, 0.891, 0.867, and 0.844, respectively. The training of light-weight neural network models based on color fundus photography for the diagnosis of common fundus diseases is not only fast but also has a significant reduction in storage size and parameter number compared with the classical classification model, and can achieve satisfactory accuracy.

Iterative image reconstruction with polar coordinate discretized system matrix for optical CT radiochromic gel dosimetry.

Gel dosimeters are a potential tool for measuring the complex dose distributions that characterize modern radiotherapy. A prototype tabletop solid-tank fan-beam optical CT scanner for readout of gel dosimeters was recently developed. This scanner does not have a straight raypath from source to detector, thus images cannot be reconstructed using filtered backprojection (FBP) and iterative techniques are required. Iterative image reconstruction requires a system matrix that describes the geometry of the imaging system. Stored system matrices can become immensely large, making them impractical for storage on a typical desktopcomputer. Here we develop a method to reduce the storage size of optical CT system matrices through use of polar coordinate discretization while accounting for the refraction in optical CTsystems. A ray tracing simulator was developed to track the path of light rays as they traverse the different mediums of the optical CT scanner. Cartesian coordinate discretized system matrices (CCDSMs) and polar coordinate discretized system matrices (PCDSMs) were generated by discretizing the reconstruction area of the optical CT scanner into a Cartesian pixel grid and a polar coordinate pixel grid, respectively. The length of each ray through each pixel was calculated and used to populate the system matrices. To ensure equal weighting during iterative reconstruction, the radial rings of PCDSMs were asymmetrically spaced such that the area of each polar pixel was constant. Two clinical phantoms and several synthetic phantoms were produced and used to evaluate the reconstruction techniques under known conditions. Reconstructed images were analyzed in terms of spatial resolution, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), signal nonuniformity (SNU), and Gamma map passpercentage. A storage size reduction of 99.72% was found when comparing a PCDSM to a CCDSM with the same total number of pixels. Images reconstructed with a PCDSM were found to have superior SNR, CNR, SNU, and Gamma (1mm, 1%) pass percentage compared to those reconstructed with a CCDSM. Increasing spatial resolution in the radial direction with increasing radial distance was found in both PCDSM and CCDSM reconstructions due to the outer regions refracting light more severely. Images reconstructed with a PCDSM showed a decrease in spatial resolution in the azimuthal directions as radial distance increases, due to the widening of the polar pixels. However, this can be mitigated with only a slight increase in storage size by increasing the number of projections. A loss of spatial resolution in the radial direction within 5mm radially from center was found when reconstructing with a PCDSM, due to the large innermost pixels. However, this was remedied by increasing the number of radial rings within the PCDSM, yielding radial spatial resolution on par with images reconstructed with a CCDSM and a storage size reduction of 99.26%. Discretizing the image pixel elements in polar coordinates achieved a system matrix storage size reduction of 99.26% with only minimal reduction in the imagequality.

TAB: Unified and Optimized Ternary, Binary, and Mixed-precision Neural Network Inference on the Edge

Ternary Neural Networks (TNNs) and mixed-precision Ternary Binary Networks (TBNs) have demonstrated higher accuracy compared to Binary Neural Networks (BNNs) while providing fast, low-power, and memory-efficient inference. Related works have improved the accuracy of TNNs and TBNs, but overlooked their optimizations on CPU and GPU platforms. First, there is no unified encoding for the binary and ternary values in TNNs and TBNs. Second, existing works store the 2-bit quantized data sequentially in 32/64-bit integers, resulting in bit-extraction overhead. Last, adopting standard 2-bit multiplications for ternary values leads to a complex computation pipeline, and efficient mixed-precision multiplication between ternary and binary values is unavailable. In this article, we propose TAB as a unified and optimized inference method for ternary, binary, and mixed-precision neural networks. TAB includes unified value representation, efficient data storage scheme and novel bitwise dot product pipelines on CPU/GPU platforms. We adopt signed integers for consistent value representation across binary and ternary values. We introduce a bitwidth-last data format that stores the first and second bits of the ternary values separately to remove the bit extraction overhead. We design the ternary and binary bitwise dot product pipelines based on Gated-XOR using up to 40% fewer operations than State-Of-The-Art (SOTA) methods. Theoretical speedup analysis shows that our proposed TAB-TNN is 2.3× fast as the SOTA ternary method RTN, 9.8× fast as 8-bit integer quantization (INT8), and 39.4× fast as 32-bit full-precision convolution (FP32). Experiment results on CPU and GPU platforms show that our TAB-TNN has achieved up to 34.6× speedup and 16× storage size reduction compared with FP32 layers. TBN, Binary-activation Ternary-weight Network (BTN), and BNN in TAB are up to 40.7×, 56.2×, and 72.2× as fast as FP32. TAB-TNN is up to 70.1% faster and 12.8% more power-efficient than RTN on Darknet-19 while keeping the same accuracy. TAB is open source as a PyTorch Extension 1 for easy integration with existing CNN models.

X-FIST: Extended flood index for efficient similarity search in massive trajectory dataset

Similarity search tasks in big trajectory datasets often require tree-based indices to shorten the query time through early pruning of dissimilar trajectories early. However, tree-based indices have been outperformed by the learned index in skewed-distribution datasets of multidimensional point experimentally. The learned index performed faster because of its data distribution awareness and machine learning model-based prediction. Directly applying learned index to trajectories can lead to inefficient query performance due to repeating range queries according to the query trajectory length. Thus, we develop X-FIST, an extended Flood index to learn the Minimum Bounding Region of the trajectories and their sub-trajectories. In similarity search, X-FIST prunes dissimilar trajectories effectively independent to the query trajectory length. If the trajectory similarity distance function changes, X-FIST does not need to train new models of its Flood index. The experimental results on three real-world trajectory datasets demonstrate that our approach shortened query time in every distance function and produced better storage size reduction than the tree-based index and direct approach of learned index.

AN OPTIMAL STORAGE ORGANIZATION FOR BLOCKCHAIN-BASED PUBLIC HEALTHCARE SYSTEM

The Blockchain technology nowadays is one of the main trends in IT industry. It has robust structure and creates trustless environment, with extensive data record liveness and accessibility due to its decentralized structure. The applications build on a Blockchain platform have potential to be user-centric with high level of synchronization between the participants. Implementation in a healthcare system can enable development of a user-centric solution for access right management. This will incentivize the patients to participate in the system, but also motivate companies (e.g. pharmaceutical and insurance companies) to shift the business strategies. This paper presents anovel approach for Blockchain-based Healthcare System (BloHeS) that introduces novel record and database organization and significantly decreases database load compared to the classic Blockchain approach. Initial findings and simulation results confirms the storage size reduction, by keeping the robustness of the system.network organization

Compressed Segmented Beat Modulation Method using Discrete Cosine Transform.

Currently used 24-hour electrocardiogram (ECG) monitors have been shown to skip detecting arrhythmias that may not occur frequently or during standardized ECG test. Hence, online ECG processing and wearable sensing applications have been becoming increasingly popular in the past few years to solve a continuous and long-term ECG monitoring problem. With the increase in the usage of online platforms and wearable devices, there arises a need for increased storage capacity to store and transmit lengthy ECG recordings, offline and over the cloud for continuous monitoring by clinicians. In this work, a discrete cosine transform (DCT) compressed segmented beat modulation method (SBMM) is proposed and its applicability in case of ambulatory ECG monitoring is tested using Massachusetts Institute of Technology-Beth Israel Deaconess Medical Center (MIT-BIH) ECG Compression Test Database containing Holter tape normal sinus rhythm ECG recordings. The method is evaluated using signal-to-noise (SNR) and compression ratio (CR) considering varying levels of signal energy in the reconstructed ECG signal. For denoising, an average SNR of 4.56 dB was achieved representing an average overall decline of 1.68 dBs (37.9%) as compared to the uncompressed signal processing while 95 % of signal energy is intact and quantized at 6 bits for signal storage (CR=2) compared to the original 12 bits, hence resulting in 50% reduction in storage size.

A Scalable Data Integrity Mechanism Based on Provable Data Possession and JARs

Cloud storage as a service provides high scalability and availability as per need of user, without large investment on infrastructure. However, data security risks, such as confidentiality, privacy, and integrity of the outsourced data are associated with the cloud-computing model. Over the year’s techniques such as, remote data checking (RDC), data integrity protection (DIP), provable data possession (PDP), proof of storage (POS), and proof of retrievability (POR) have been devised to frequently and securely check the integrity of outsourced data. In this paper, we improve the efficiency of PDP scheme, in terms of computation, storage, and communication cost for large data archives. By utilizing the capabilities of JAR and ZIP technology, the cost of searching the metadata in proof generation process is reduced from O(n) toO(1). Moreover, due to direct access to metadata, disk I/O cost is reduced and resulting in 50 to 60 time faster proof generation for large datasets. Furthermore, our proposed scheme achieved 50% reduction in storage size of data and respective metadata that result in providing storage and communication efficiency.

OPTICAL CHARACTER RECOGNITION USING ARTIFICIAL NEURAL NETWORK

Optical character recognition refers to the process of translat ing images of hand-written, typewritten, or printed text into a format understood by machines for the purpose of editing, indexing/searching, and a reduction in storage size. Optical character recognition is the mechanical or electronic translation of images of handwritten, typewritten or printed text into machine-editable text. Artificial neural networks are commonly used to perform character recognition due to their high noise tolerance. In this paper, an Optical character recognition based on Artificial Neural Networks (ANNs). The ANN is trained using the Back Propagation algorithm.

A Resource Efficient Big Data Analysis Method for the Social Sciences: The Case of Global IP Activity

This paper presents a novel and efficient way of analysing big datasets used in social science research. We provide and demonstrate a way to deal with such datasets without the need for high performance distributed computational facilities. Using an Internet census dataset and with the help of freely available tools and programming libraries, we visualize global IP activity in a spatial and time dimension. We observe a considerable reduction in storage size of our dataset coupled with a faster processing time.

Effect of connecting rain barrels on the storage size reduction

AbstractIn this article, the possibility of sharing rain barrels and the potential benefit of reducing storage size through physical and non‐physical connections of rain barrels in a community are investigated. Using the concepts of homogeneous/heterogeneous users in rainwater harvesting systems (RWHS), two simple cases of a community composed of four prospective users are examined. The first is performed with the users who have the same mean and variance in water demands (homogeneous users), and the second is with the users with different means and variances (heterogeneous users). To take account for the rainfall characteristics in different places, historical records from six cities in the USA are used for storage–reliability–yield analysis. The result indicates that required total storage can be reduced by connecting multiple rain barrels. In addition, a significant difference is found between homogeneous and heterogeneous user groups. Homogeneous users do not achieve a substantial benefit from connecting their rain barrels; these users may even be disadvantaged by sharing. In contrast, heterogeneous users receive benefit by reducing the total required storage. Most benefit is expected between users with maximum difference in mean water demands. The reduction in storage size was as considerable as 37% in this study. The quantity of storage reduction depends on locations and target reliabilities. Knowledge of the benefits and limitations of rain barrel connections can improve RWHS performance through ability to customize a network plan for individual users. Copyright © 2011 John Wiley & Sons, Ltd.