RESUMO
Previous knowledge distillation (KD) methods mostly focus on compressing network architectures, which is not thorough enough in deployment as some costs like transmission bandwidth and imaging equipment are related to the image size. Therefore, we propose Pixel Distillation that extends knowledge distillation into the input level while simultaneously breaking architecture constraints. Such a scheme can achieve flexible cost control for deployment, as it allows the system to adjust both network architecture and image quality according to the overall requirement of resources. Specifically, we first propose an input spatial representation distillation (ISRD) mechanism to transfer spatial knowledge from large images to student's input module, which can facilitate stable knowledge transfer between CNN and ViT. Then, a Teacher-Assistant-Student (TAS) framework is further established to disentangle pixel distillation into the model compression stage and input compression stage, which significantly reduces the overall complexity of pixel distillation and the difficulty of distilling intermediate knowledge. Finally, we adapt pixel distillation to object detection via an aligned feature for preservation (AFP) strategy for TAS, which aligns output dimensions of detectors at each stage by manipulating features and anchors of the assistant. Comprehensive experiments on image classification and object detection demonstrate the effectiveness of our method.
RESUMO
Traditional cathode recycling methods have become outdated amid growing concerns for high-value output and environmental friendliness in spent Li-ion battery (LIB) recycling. Our study presents a closed-loop approach that involves selective sulfurization roasting, water leaching, and regeneration, efficiently transforming spent ternary Li batteries (i.e., NCM) into high-performance cathode materials. By combining experimental investigations with density functional theory (DFT) calculations, we elucidate the mechanisms within the NCM-C-S roasting system, providing a theoretical foundation for selective sulfidation. Utilizing in situ X-ray diffraction techniques and a series of consecutive experiments, the study meticulously tracks the evolution of regenerating cathode materials that use transition metal sulfides as their primary raw materials. The Li-rich regenerated NCM exhibits exceptional electrochemical performance, including long-term cycling, high-rate capabilities, reversibility, and stability. The closed-loop approach highlights the sustainability and environmental friendliness of this recycling process, with potential applications in other cathode materials, such as LiCoO2 and LiMn2O4. Compared with traditional methods, this short process approach avoids the complexity of leaching, solvent extraction, and reverse extraction, significantly increasing metal utilization and Li recovery rates while reducing pollution and resource waste.
Assuntos
Lítio , Metais , Fontes de Energia Elétrica , Eletrodos , Reciclagem , ÍonsRESUMO
The facile recycling of spent lithium-ion batteries (LIBs) has attracted considerable attention because of its great importance to environmental protection and resource utilization. A novel process is developed for cyclic utilization of spent LiNixCoyMnzO2 (NCM) batteries. The spent NCM was converted into water-soluble Li2CO3, acid-dissolved MnO, and nickel-cobalt sulfides through selective sulfidation, based on roasting condition optimization and thermodynamic calculation. More than 98 % of lithium is extracted preferentially from calcined NCM through water leaching, and over 99 % of manganese is extracted selectively from water leaching residue with H2SO4 solution of 0.4 mol/L in the absence of additional reductant. The nickel and cobalt sulfides were concentrated into the leaching residue without metal impurities. The obtained Li2CO3, MnSO4, and nickel-cobalt sulfides can be regenerated as new NCM, showing good electrochemical performance, and its discharge capacity is 169.8 mAh/g at 0.2C. After 100 cycles at 0.2C, the discharge specific capacity can still be maintained at 143.24 mAh/g, and its capacity retention ratio is as high as 92 %. An environmental assessment and economic evaluation indicate that the process is an economical and eco-friendly approach for green recycling of spent LIBs.
Assuntos
Lítio , Níquel , Cobalto , Fontes de Energia Elétrica , Reciclagem , SulfetosRESUMO
The facile recycling of spent lithium-ion batteries (LIBs) has attracted much attention because of its great significance to the environmental protection and resource utilization. Hydrometallurgical process is the most common method for recycling spent LIBs, but it is difficult to economically recover spent LiFePO4 batteries, because of the complicated metal separation process and low added value of its products. Herein, a novel and facile approach has been developed to achieve the direct regeneration of LiFePO4 from spent LIBs. By employing a flotation process after effective pyrolysis, it is found that 91.57% of LiFePO4 can be recovered from spent LIBs. Different surface hydrophobicity of cathode and anode active materials could be achieved via the selective adsorption of causticized soluble starch on the surfaces of spent LiFePO4, which effectively enhances the separation performance in flotation process. The recovered LiFePO4 barely contains metal impurities, which can be directly regenerated as new LiFePO4 materials with the first discharge capacity of 161.37 mAh/g, and their capacity retention is as high as 97.53% after 100 cycles at 0.2C. A technology assessment and economic evaluation indicate the developed regeneration approach of LiFePO4 is environmentally and economically feasible, which avoids the complex element separation process and achieves the facile recycling of spent LiFePO4.
RESUMO
Accurate assessment of renal histopathology is crucial for the clinical management of patients with lupus nephritis (LN). However, the current classification system has poor interpathologist agreement. This paper proposes a deep convolutional neural network (CNN)-based system that detects and classifies glomerular pathological findings in LN. A dataset of 349 renal biopsy whole-slide images (WSIs) (163 patients with LN, periodic acid-Schiff stain, 3906 glomeruli) annotated by three expert nephropathologists was used. The CNN models YOLOv4 and VGG16 were employed to localise the glomeruli and classify glomerular lesions (slight/severe impairments or sclerotic lesions). An additional 321 unannotated WSIs from 161 patients were used for performance evaluation at the per-patient kidney level. The proposed model achieved an accuracy of 0.951 and Cohen's kappa of 0.932 (95% CI 0.915-0.949) for the entire test set for classifying the glomerular lesions. For multiclass detection at the glomerular level, the mean average precision of the CNN was 0.807, with 'slight' and 'severe' glomerular lesions being easily identified (F1: 0.924 and 0.952, respectively). At the per-patient kidney level, the model achieved a high agreement with nephropathologist (linear weighted kappa: 0.855, 95% CI: 0.795-0.916, p < 0.001; quadratic weighted kappa: 0.906, 95% CI: 0.873-0.938, p < 0.001). The results suggest that deep learning is a feasible assistive tool for the objective and automatic assessment of pathological LN lesions.
RESUMO
In recent years, predicting the saccadic scanpaths of humans has become a new trend in the field of visual attention modeling. Given various saccadic algorithms, determining how to evaluate their ability to model a dynamic saccade has become an important yet understudied issue. To our best knowledge, existing metrics for evaluating saccadic prediction models are often heuristically designed, which may produce results that are inconsistent with human subjective assessment. To this end, we first construct a subjective database by collecting the assessments on 5,000 pairs of scanpaths from ten subjects. Based on this database, we can compare different metrics according to their consistency with human visual perception. In addition, we also propose a data-driven metric to measure scanpath similarity based on the human subjective comparison. To achieve this goal, we employ a long short-term memory (LSTM) network to learn the inference from the relationship of encoded scanpaths to a binary measurement. Experimental results have demonstrated that the LSTM-based metric outperforms other existing metrics. Moreover, we believe the constructed database can be used as a benchmark to inspire more insights for future metric selection.
Assuntos
Algoritmos , Movimentos Sacádicos , Bases de Dados Factuais , Humanos , Redes Neurais de Computação , Percepção VisualRESUMO
OBJECTIVE: Various studies in the brain mapping field have demonstrated that there exist multiple concurrent functional networks that are spatially overlapped and interacting with each other during specific task performance to jointly realize the total brain function. Assessing such spatial overlap patterns of functional networks (SOPFNs) based on functional magnetic resonance imaging (fMRI) has thus received increasing interest for brain function studies. However, there are still two crucial issues to be addressed. First, the SOPFNs are assessed over the entire fMRI scan assuming the temporal stationarity, while possibly time-dependent dynamics of the SOPFNs is not sufficiently explored. Second, the SOPFNs are assessed within individual subjects, while group-wise consistency of the SOPFNs is largely unknown. METHODS: To address the two issues, we propose a novel computational framework of group-wise sparse representation of whole-brain fMRI temporal segments to assess the temporal dynamic spatial patterns of SOPFNs that are consistent across different subjects. RESULTS: Experimental results based on the recently publicly released Human Connectome Project grayordinate task fMRI data demonstrate that meaningful SOPFNs exhibiting dynamic spatial patterns across different time periods are effectively and robustly identified based on the reconstructed concurrent functional networks via the proposed framework. Specifically, those SOPFNs locate significantly more on gyral regions than on sulcal regions across different time periods. CONCLUSION: These results reveal novel functional architecture of cortical gyri and sulci. SIGNIFICANCE: Moreover, these results help better understand functional dynamics mechanisms of cerebral cortex in the future.
Assuntos
Mapeamento Encefálico/métodos , Córtex Cerebral/diagnóstico por imagem , Córtex Cerebral/fisiologia , Imageamento por Ressonância Magnética/métodos , Algoritmos , Humanos , Processamento de Imagem Assistida por Computador , Rede Nervosa/fisiologia , Aprendizado de Máquina SupervisionadoRESUMO
Triple negative breast cancers (TNBCs) are highly heterogeneous and aggressive without targeted treatment. Here, we aim to systematically dissect TNBCs from a prognosis point of view by building a subnetwork atlas for TNBC prognosis through integrating multi-dimensional cancer genomics data from The Cancer Genome Atlas (TCGA) project and the interactome data from three different interaction networks. The subnetworks are represented as the protein-protein interaction modules perturbed by multiple genetic and epigenetic interacting mechanisms contributing to patient survival. Predictive power of these subnetwork-derived prognostic models is evaluated using Monte Carlo cross-validation and the concordance index (C-index). We uncover subnetwork biomarkers of low oncogenic GTPase activity, low ubiquitin/proteasome degradation, effective protection from oxidative damage, and tightly immune response are linked to better prognosis. Such a systematic approach to integrate massive amount of cancer genomics data into clinical practice for TNBC prognosis can effectively dissect the molecular mechanisms underlying TNBC patient outcomes and provide potential opportunities to optimize treatment and develop therapeutics.