Semantic Regions Research Articles

Crop Classification from Drone Imagery Based on Lightweight Semantic Segmentation Methods

Technological advances have dramatically improved precision agriculture, and accurate crop classification is a key aspect of precision agriculture (PA). The flexibility and real-time nature of UAVs have led them to become an important tool for acquiring agricultural data and enabling precise crop classification. Currently, crop identification relies heavily on complex high-precision models that often struggle to provide real-time performance. Research on lightweight models specifically for crop classification is also limited. In this paper, we propose a crop classification method based on UAV visible-light images based on PP-LiteSeg, a lightweight model proposed by Baidu. To improve the accuracy, a pyramid pooling module is designed in this paper, which integrates adaptive mean pooling and CSPC (Convolutional Spatial Pyramid Pooling) techniques to handle high-resolution features. In addition, a sparse self-attention mechanism is employed to help the model pay more attention to locally important semantic regions in the image. The combination of adaptive average pooling and the sparse self-attention mechanism can better handle different levels of contextual information. To train the model, a new dataset based on UAV visible-light images including nine categories such as rice, soybean, red bean, wheat, corn, poplar, etc., with a time span of two years was created for accurate crop classification. The experimental results show that the improved model outperforms other models in terms of accuracy and prediction performance, with a MIoU (mean intersection ratio joint) of 94.79%, which is 2.79% better than the original model. Based on the UAV RGB images demonstrated in this paper, the improved model achieves a better balance between real-time performance and accuracy. In conclusion, the method effectively utilizes UAV RGB data and lightweight deep semantic segmentation models to provide valuable insights for crop classification and UAV field monitoring.

Semi-supervised image semantic segmentation method with semantic regions patching and uncertainty-guided loss

Semi-supervised image semantic segmentation method with semantic regions patching and uncertainty-guided loss

Robust Model Watermarking for Image Processing Networks via Structure Consistency.

The intellectual property of deep networks can be easily "stolen" by surrogate model attack. There has been significant progress in protecting the model IP in classification tasks. However, little attention has been devoted to the protection of image processing models. By utilizing consistent invisible spatial watermarks, the work (Zhang et al. 2020) first considered model watermarking for deep image processing networks and demonstrated its efficacy in many downstream tasks. Its success depends on the hypothesis that if a consistent watermark exists in all prediction outputs, that watermark will be learned into the attacker's surrogate model. However, when the attacker uses common data augmentation attacks (e.g., rotate, crop, and resize) during surrogate model training, it will fail because the underlying watermark consistency is destroyed. To mitigate this issue, we propose a new watermarking methodology, "structure consistency", based on which a new deep structure-aligned model watermarking algorithm is designed. Specifically, the embedded watermarks are designed to be aligned with physically consistent image structures, such as edges or semantic regions. Experiments demonstrate that our method is more robust than the baseline in resisting data augmentation attacks. Besides that, we test the generalization ability and robustness of our method to a broader range of adaptive attacks.

Explainable machine learning for assessing upper respiratory tract of racehorses from endoscopy videos

Laryngeal hemiplegia (LH) is a major upper respiratory tract (URT) complication in racehorses. Endoscopy imaging of horse throat is a gold standard for URT assessment. However, current manual assessment faces several challenges, stemming from the poor quality of endoscopy videos and subjectivity of manual grading. To overcome such limitations, we propose an explainable machine learning (ML)-based solution for efficient URT assessment. Specifically, a cascaded YOLOv8 architecture is utilized to segment the key semantic regions and landmarks per frame. Several spatiotemporal features are then extracted from key landmarks points and fed to a decision tree (DT) model to classify LH as Grade 1,2,3 or 4 denoting absence of LH, mild, moderate, and severe LH, respectively. The proposed method, validated through 5-fold cross-validation on 107 videos, showed promising performance in classifying different LH grades with 100%, 91.18%, 94.74% and 100% sensitivity values for Grade 1 to 4, respectively. Further validation on an external dataset of 72 cases confirmed its generalization capability with 90%, 80.95%, 100%, and 100% sensitivity values for Grade 1 to 4, respectively. We introduced several explainability related assessment functions, including: (i) visualization of YOLOv8 output to detect landmark estimation errors which can affect the final classification, (ii) time-series visualization to assess video quality, and (iii) backtracking of the DT output to identify borderline cases. We incorporated domain knowledge (e.g., veterinarian diagnostic procedures) into the proposed ML framework. This provides an assistive tool with clinical-relevance and explainability that can ease and speed up the URT assessment by veterinarians.

Out-of-Distribution Detection with Virtual Outlier Smoothing

AbstractDetecting out-of-distribution (OOD) inputs plays a crucial role in guaranteeing the reliability of deep neural networks (DNNs) when deployed in real-world scenarios. However, DNNs typically exhibit overconfidence in OOD samples, which is attributed to the similarity in patterns between OOD and in-distribution (ID) samples. To mitigate this overconfidence, advanced approaches suggest the incorporation of auxiliary OOD samples during model training, where the outliers are assigned with an equal likelihood of belonging to any category. However, identifying outliers that share patterns with ID samples poses a significant challenge. To address the challenge, we propose a novel method, Virtual Outlier Smoothing (VOSo), which constructs auxiliary outliers using ID samples, thereby eliminating the need to search for OOD samples. Specifically, VOSo creates these virtual outliers by perturbing the semantic regions of ID samples and infusing patterns from other ID samples. For instance, a virtual outlier might consist of a cat’s face with a dog’s nose, where the cat’s face serves as the semantic feature for model prediction. Meanwhile, VOSo adjusts the labels of virtual OOD samples based on the extent of semantic region perturbation, aligning with the notion that virtual outliers may contain ID patterns. Extensive experiments are conducted on diverse OOD detection benchmarks, demonstrating the effectiveness of the proposed VOSo. Our code will be available at https://github.com/junz-debug/VOSo.

Semantic-aware hyper-space deformable neural radiance fields for facial avatar reconstruction

High-fidelity facial avatar reconstruction from monocular videos is a prominent research problem in computer graphics and computer vision. Recent advancements in the Neural Radiance Field (NeRF) have demonstrated remarkable proficiency in rendering novel views and garnered attention for its potential in facial avatar reconstruction. However, previous methodologies have overlooked the complex motion dynamics present across the head, torso, and intricate facial features. Additionally, a deficiency exists in a generalized NeRF-based framework for facial avatar reconstruction adaptable to either 3DMM coefficients or audio input. To tackle these challenges, we propose an innovative framework that leverages semantic-aware hyper-space deformable NeRF, facilitating the reconstruction of high-fidelity facial avatars from either 3DMM coefficients or audio features. Our framework effectively addresses both localized facial movements and broader head and torso motions through semantic guidance and a unified hyper-space deformation module. Specifically, we adopt a dynamic weighted ray sampling strategy to allocate varying degrees of attention to distinct semantic regions, enhancing the deformable NeRF framework with semantic guidance to capture fine-grained details across diverse facial regions. Moreover, we introduce a hyper-space deformation module that enables the transformation of observation space coordinates into canonical hyper-space coordinates, allowing for the learning of natural facial deformation and head-torso movements. Extensive experiments validate the superiority of our framework over existing state-of-the-art methods, demonstrating its effectiveness in producing realistic and expressive facial avatars. Our code is available at https://github.com/jematy/SAHS-Deformable-Nerf.

Semantic-aware Message Broadcasting for Efficient Unsupervised Domain Adaptation.

Vision transformer has demonstrated great potential in abundant vision tasks. However, it also inevitably suffers from poor generalization capability when the distribution shift occurs in testing (i.e., out-of-distribution data). To mitigate this issue, we propose a novel method, Semantic-aware Message Broadcasting (SAMB), which enables more informative and flexible feature alignment for unsupervised domain adaptation (UDA). Particularly, we study the attention module in the vision transformer and notice that the alignment space using one global class token lacks enough flexibility, where it interacts information with all image tokens in the same manner but ignores the rich semantics of different regions. In this paper, we aim to improve the richness of the alignment features by enabling semantic-aware adaptive message broadcasting. Particularly, we introduce a group of learned group tokens as nodes to aggregate the global information from all image tokens, but encourage different group tokens to adaptively focus on the message broadcasting to different semantic regions. In this way, our message broadcasting encourages the group tokens to learn more informative and diverse information for effective domain alignment. Moreover, we systematically study the effects of adversarial-based feature alignment (ADA) and pseudo-label based self-training (PST) on UDA. We find that one simple two-stage training strategy with the cooperation of ADA and PST can further improve the adaptation capability of the vision transformer. Extensive experiments on DomainNet, OfficeHome, and VisDA-2017 demonstrate the effectiveness of our methods for UDA.

PatchNet: Maximize the Exploration of Congeneric Semantics for Weakly Supervised Semantic Segmentation.

With the increase in the number of image data and the lack of corresponding labels, weakly supervised learning has drawn a lot of attention recently in computer vision tasks, especially in the fine-grained semantic segmentation problem. To alleviate human efforts from expensive pixel-by-pixel annotations, our method focuses on weakly supervised semantic segmentation (WSSS) with image-level labels, which are much easier to obtain. As a considerable gap exists between pixel-level segmentation and image-level labels, how to reflect the image-level semantic information on each pixel is an important question. To explore the congeneric semantic regions from the same class to the maximum, we construct the patch-level semantic augmentation network (PatchNet) based on the self-detected patches from different images that contain the same class labels. Patches can frame the objects as much as possible and include as little background as possible. The patch-level semantic augmentation network that is established with patches as the nodes can maximize the mutual learning of similar objects. We regard the embedding vectors of patches as nodes and use a transformer-based complementary learning module to construct weighted edges according to the embedding similarity between different nodes. Moreover, to better supplement semantic information, we propose softcomplementary loss functions matched with the whole network structure. We conduct experiments on the popular PASCAL VOC 2012 and MS COCO 2014 benchmarks, and our model yields the state-of-the-art performance.

Semantic attention guided low-light image enhancement with multi-scale perception

Low-light environments often lead to complex degradation of captured images. However, most deep learning-based image enhancement methods for low-light conditions only learn a single-channel mapping relationship between the input image in low-light conditions and the desired image in normal light without considering semantic priors. This may cause the network to deviate from the original color of the region. In addition, deep network architectures are not suitable for low-light image recovery due to low pixel values. To address these issues, we propose a novel network called SAGNet. It consists of two branches:the main branch extracts global enhancement features at the level of the original image, and the other branch introduces semantic information through region-based feature learning and learns local enhancement features for semantic regions with multi-level perception to maintain color consistency. The extracted features are merged with the global enhancement features for semantic consistency and visualization. We also propose an unsupervised loss function to improve the network’s adaptability to general scenes and reduce the effect of sparse datasets. Extensive experiments and ablation studies show that SAGNet maintains color accuracy better in all cases and keeps natural luminance consistency across the semantic range.

Functional integration and segregation during semantic cognition: Evidence across age groups

Semantic cognition is underpinned by ventral anterior temporal lobe (vATL) which encodes knowledge representations and inferior frontal gyrus (IFG), which controls activation of knowledge based on the needs of the current context. This core semantic network has been validated in substantial empirical findings in the past. However, it remains unclear how these core semantic areas dynamically communicate with each other, and with other neural networks, to achieve successful semantic processing. Here, we investigated this question by testing functional connectivity in the core semantic network during semantic tasks and whether these connections were affected by cognitive ageing. Compared to a non-semantic task, semantic tasks increased the connectivity between left and right IFGs, indicating a bilateral semantic control system. Strengthened connectivity was also found between left IFG and left vATL, and this effect was stronger in the young group. At a whole-brain scale, IFG and vATL increased their coupling with multiple-demand regions during semantic tasks, even though these areas were deactivated relative to non-semantic tasks. This suggests that the domain-general executive network contributes to semantic processing. In contrast, IFG and vATL decreased their interaction with default mode network (DMN) areas during semantic tasks, even though these areas were positively activated by the task. This suggests that DMN areas do not contribute to all semantic tasks: their activation may sometimes reflect automatic retrieval of task-irrelevant memories and associations. Taken together, our study characterizes a dynamic connectivity mechanism supporting semantic cognition within and beyond core semantic regions.

Long-term motor training enhances functional connectivity between semantic and motor regions in an effector-specific manner: evidence from elite female football athletes.

The relation between the action verb semantic processing and sensorimotor experience remains controversial. In this study, we examined whether plasticity changes in brain are specifically related to semantic processing of foot action verbs when long-term motor training is mainly aimed at the foot. To address this question, we acquired resting-state functional magnetic resonance imaging scans and behavioral data from a verb two-choice task from female expertise football players and football novices. We compared the resting-state functional connectivity (rsFC) differences between experts and novices using motor execution regions and general semantic regions (left anterior temporal lobe, lATL) as seed, and explored the neural correlates of behavioral performance. Here, the drift rate (v) parameter of the drift diffusion model (DDM) was used to capture the semantic processing capability. We found experts showed increased correlation between lATL subregions and important brain regions for motor processing, including supplementary motor area (SMA), bilateral paracentral lobule (PL), superior parietal lobule and inferior parietal lobule, in contrast to novices. Further predictive model analysis showed the FC found in rsFC analysis can significantly predict drift rate of foot action verb in both experts and novices, but not drift rate of hand action verb. Our findings therefore establish a connection between effector-related semantic processing and the plasticity changes in brain functional connectivity, attributable to long-term foot-related motor training. This provides evidence supporting the view that semantic processing is fundamentally rooted in the sensorimotor system.

Person re-identification by utilizing hierarchical spatial relation reasoning

Deep learning-based methods have become increasingly popular in person re-identification (re-id) as they enhance the discriminative features while suppressing extraneous or irrelevant details. However, these methods face a potential limitation in overlooking the relationships between semantic regions focusing solely on salient features or pixel-level matching information. As a result, the model performance becomes susceptible to adverse factors such as pose variations, misalignment, and background clutter. To overcome this issue, we propose a novel deep relational learning network (DRLNet) that learns relationship knowledge between semantic regions and “local–global” information to improve discriminative feature learning effectively. We manually establish the relation between local regions using a simple partition strategy that captures global structural information and improves relation-aware attention learning. Dividing images into uniform partitions allows each part-level feature to represent itself and the entire image area, facilitating the learning of more refined and discerning relationships. Our approach's efficacy is evaluated on three datasets, where experimental results exhibit its superiority. Significantly, our approach achieves a new state-of-the-art Rank-1 accuracy of 84.7%, 80.9%, and 83.5% on CUHK03 (Labeled), CUHK03 (Detected), and MSMT17 datasets, respectively, outperforming current state-of-the-art methods. In conclusion, our proposed DRLNet demonstrates its potential as a highly effective solution to overcoming the limitations of deep learning-based methods by providing superior means of learning discriminative features while incorporating relationships between semantic regions.

Verbal semantic expertise is associated with reduced functional connectivity between left and right anterior temporal lobes.

The left and right anterior temporal lobes (ATLs) encode semantic representations. They show graded hemispheric specialization in function, with the left ATL contributing preferentially to verbal semantic processing. We investigated the cognitive correlates of this organization, using resting-state functional connectivity as a measure of functional segregation between ATLs. We analyzed two independent resting-state fMRI datasets (n= 86 and n= 642) in which participants' verbal semantic expertise was measured using vocabulary tests. In both datasets, people with more advanced verbal semantic knowledge showed weaker functional connectivity between left and right ventral ATLs. This effect was highly specific. It was not observed for within-hemisphere connections between semantic regions (ventral ATL and inferior frontal gyrus (IFG), though it was found for left-right IFG connectivity in one dataset). Effects were not found for tasks probing semantic control, nonsemantic cognition, or face recognition. Our results suggest that hemispheric specialization in the ATLs is not an innate property but rather emerges as people develop highly detailed verbal semantic representations. We speculate that this effect is a consequence of the left ATL's greater connectivity with left-lateralized written word recognition regions, which causes it to preferentially represent meaning for advanced vocabulary acquired primarily through reading.

SPNet: Semantic preserving network with semantic constraint and non-semantic calibration for color constancy

Recent methods introduce semantics obtained from pre-trained classification models into color constancy to guide the model in learning object-color mapping, thereby improving the illumination estimation ability. However, the task discrepancy between classification and color constancy leads to a representation learning gap, resulting in semantic loss during training. Moreover, semantic-based methods emphasize semantic regions, leading to the neglect of clues contained in non-semantic regions. Overall, semantic-based methods face the challenge of underutilizing critical information. To address this problem, we propose a Semantic Preserving Network (SPNet). We first design a Semantic Constraint Module (SCM) in SPNet. SCM provides semantic constraints to ensure uninterrupted semantic knowledge transfer during training, which prevents semantic loss. Additionally, we further propose an Auxiliary Calibration Module (ACM). ACM explores background regions with a restricted range of innate colors. The high consistency property of their color contributes to the calibration of illumination colors. Meanwhile, ACM recalibrates local–global consistency to avoid large estimation bias in other background regions where illumination clues are insufficient. Extensive experiments on the benchmark datasets show that our method achieves superior performance without extra inference consumption.

Score-based Graph Learning for Urban Flow Prediction

Accurate urban flow prediction (UFP) is crucial for a range of smart city applications such as traffic management, urban planning, and risk assessment. To capture the intrinsic characteristics of urban flow, recent efforts have utilized spatial and temporal graph neural networks to deal with the complex dependence between the traffic in adjacent areas. However, existing graph neural network based approaches suffer from several critical drawbacks, including improper graph representation of urban traffic data, lack of semantic correlation modeling among graph nodes, and coarse-grained exploitation of external factors. To address these issues, we propose DiffUFP , a novel probabilistic graph-based framework for UFP. DiffUFP consists of two key designs: (1) a semantic region dynamic extraction method that effectively captures the underlying traffic network topology, and (2) a conditional denoising score-based adjacency matrix generator that takes spatial, temporal, and external factors into account when constructing the adjacency matrix rather than simply concatenation in existing studies. Extensive experiments conducted on real-world datasets demonstrate the superiority of DiffUFP over the state-of-the-art UFP models and the effect of the two specific modules.

Conditioning diffusion models via attributes and semantic masks for face generation

Deep generative models have shown impressive results in generating realistic images of faces. GANs managed to generate high-quality, high-fidelity images when conditioned on semantic masks, but they still lack the ability to diversify their output. Diffusion models partially solve this problem and are able to generate diverse samples given the same condition. This paper introduces a novel strategy for enhancing diffusion models through multi-conditioning, harnessing cross-attention mechanisms to utilize multiple feature sets, ultimately enabling the generation of high-quality and controllable images. The proposed method extends previous approaches by introducing conditioning on both attributes and semantic masks, ensuring finer control over the generated face images. In order to improve the training time and the generation quality, the impact of applying perceptual-focused loss weighting into the latent space instead of the pixel space is also investigated. The proposed solution has been evaluated on the CelebA-HQ dataset, and it can generate realistic and diverse samples while allowing for fine-grained control over multiple attributes and semantic regions. Experiments on the DeepFashion dataset have also been performed in order to analyze the capability of the proposed model to generalize to different domains. In addition, an ablation study has been conducted to evaluate the impact of different conditioning strategies on the quality and diversity of the generated images.

Development and validation of a nonverbal consensus-based semantic memory paradigm in patients with epilepsy.

Brain areas implicated in semantic memory can be damaged in patients with epilepsy (PWE). However, it is challenging to delineate semantic processing deficits from acoustic, linguistic, and other verbal aspects in current neuropsychological assessments. We developed a new Visual-based Semantic Association Task (ViSAT) to evaluate nonverbal semantic processing in PWE. The ViSAT was adapted from similar predecessors (Pyramids & Palm Trees test, PPT; Camels & Cactus Test, CCT) comprised of 100 unique trials using real-life color pictures that avoid demographic, cultural, and other potential confounds. We obtained performance data from 23 PWE participants and 24 control participants (Control), along with crowdsourced normative data from 54 Amazon Mechanical Turk (Mturk) workers. ViSAT reached a consensus >90% in 91.3% of trials compared to 83.6% in PPT and 82.9% in CCT. A deep learning model demonstrated that visual features of the stimulus images (color, shape; i.e., non-semantic) did not influence top answer choices (p = 0.577). The PWE group had lower accuracy than the Control group (p = 0.019). PWE had longer response times than the Control group in general and this was augmented for the semantic processing (trial answer) stage (both p < 0.001). This study demonstrated performance impairments in PWE that may reflect dysfunction of nonverbal semantic memory circuits, such as seizure onset zones overlapping with key semantic regions (e.g., anterior temporal lobe). The ViSAT paradigm avoids confounds, is repeatable/longitudinal, captures behavioral data, and is open-source, thus we propose it as a strong alternative for clinical and research assessment of nonverbal semantic memory.

HaLo‐NeRF: Learning Geometry‐Guided Semantics for Exploring Unconstrained Photo Collections

AbstractInternet image collections containing photos captured by crowds of photographers show promise for enabling digital exploration of large‐scale tourist landmarks. However, prior works focus primarily on geometric reconstruction and visualization, neglecting the key role of language in providing a semantic interface for navigation and fine‐grained understanding. In more constrained 3D domains, recent methods have leveraged modern vision‐and‐language models as a strong prior of 2D visual semantics. While these models display an excellent understanding of broad visual semantics, they struggle with unconstrained photo collections depicting such tourist landmarks, as they lack expert knowledge of the architectural domain and fail to exploit the geometric consistency of images capturing multiple views of such scenes. In this work, we present a localization system that connects neural representations of scenes depicting large‐scale landmarks with text describing a semantic region within the scene, by harnessing the power of SOTA vision‐and‐language models with adaptations for understanding landmark scene semantics. To bolster such models with fine‐grained knowledge, we leverage large‐scale Internet data containing images of similar landmarks along with weakly‐related textual information. Our approach is built upon the premise that images physically grounded in space can provide a powerful supervision signal for localizing new concepts, whose semantics may be unlocked from Internet textual metadata with large language models. We use correspondences between views of scenes to bootstrap spatial understanding of these semantics, providing guidance for 3D‐compatible segmentation that ultimately lifts to a volumetric scene representation. To evaluate our method, we present a new benchmark dataset containing large‐scale scenes with ground‐truth segmentations for multiple semantic concepts. Our results show that HaLo‐NeRF can accurately localize a variety of semantic concepts related to architectural landmarks, surpassing the results of other 3D models as well as strong 2D segmentation baselines. Our code and data are publicly available at https://tau‐vailab.github.io/HaLo‐NeRF/.

An Adaptive Semantic Segmentation Network for Adversarial Learning Domain Based on Low-Light Enhancement and Decoupled Generation

Nighttime semantic segmentation due to issues such as low contrast, fuzzy imaging, and low-quality annotation results in significant degradation of masks. In this paper, we introduce a domain adaptive approach for nighttime semantic segmentation that overcomes the reliance on low-light image annotations to transfer the source domain model to the target domain. On the front end, a low-light image enhancement sub-network combining lightweight deep learning with mapping curve iteration is adopted to enhance nighttime foreground contrast. In the segmentation network, the body generation and edge preservation branches are implemented to generate consistent representations within the same semantic region. Additionally, a pixel weighting strategy is embedded to increase the prediction accuracy for small targets. During the training, a discriminator is implemented to distinguish features between the source and target domains, thereby guiding the segmentation network for adversarial transfer learning. The proposed approach’s effectiveness is verified through testing on Dark Zurich, Nighttime Driving, and CityScapes, including evaluations of mIoU, PSNR, and SSIM. They confirm that our approach surpasses existing baselines in segmentation scenarios.

Dissociable Contributions of the Medial Parietal Cortex to Recognition Memory.

Human neuroimaging studies of episodic memory retrieval routinely observe the engagement of specific cortical regions beyond the medial temporal lobe. Of these, medial parietal cortex (MPC) is of particular interest given its distinct functional characteristics during different retrieval tasks. Specifically, while recognition and autobiographical recall tasks are both used to probe episodic retrieval, these paradigms consistently drive distinct spatial patterns of response within MPC. However, other studies have emphasized alternate MPC functional dissociations in terms of brain network connectivity profiles or stimulus category selectivity. As the unique contributions of MPC to episodic memory remain unclear, adjudicating between these different accounts can provide better consensus regarding MPC function. Therefore, we used a precision-neuroimaging dataset (7T functional magnetic resonance imaging) to examine how MPC regions are differentially engaged during recognition memory and how these task-related dissociations may also reflect distinct connectivity and stimulus category functional profiles. We observed interleaved, though spatially distinct, subregions of MPC where responses were sensitive to either recognition decisions or the semantic representation of stimuli. In addition, this dissociation was further accentuated by functional subregions displaying distinct profiles of connectivity with the hippocampus during task and rest. Finally, we show that recent observations of dissociable person and place selectivity within the MPC reflect category-specific responses from within identified semantic regions that are sensitive to mnemonic demands. Together, by examining precision functional mapping within individuals, these data suggest that previously distinct observations of functional dissociation within MPC conform to a common principle of organization throughout hippocampal-neocortical memory systems.