Fine-grained classification based on multi-scale pyramid convolution networks.
Ontology highlight
ABSTRACT:
SUBMITTER: Wang G
PROVIDER: S-EPMC8270455 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:Crowd counting has been widely studied by deep learning in recent years. However, due to scale variation caused by perspective distortion, crowd counting is still a challenging task. In this paper, we propose a Densely Connected Multi-scale Pyramid Network (DMPNet) for count estimation and the generation of high-quality density maps. The key component of our network is the Multi-scale Pyramid Network (MPN), which can extract multi-scale features of the crowd effectively while keeping the resolution of the input feature map and the number of channels unchanged. To increase the information transfer between the network layer, we used dense connections to connect multiple MPNs. In addition, we also designed a novel loss function, which can help our model achieve better convergence. To evaluate our method, we conducted extensive experiments on three challenging benchmark crowd counting datasets. Experimental results show that compared with the state-of-the-art algorithms, DMPNet performs well in both parameters and results. The code is available at: https://github.com/lpfworld/DMPNet.
| S-EPMC9044264 | biostudies-literature
Project description:Generative adversarial networks (GANs) have been recently applied to medical imaging on different modalities (MRI, CT, X-ray, etc). However there are not many applications on ultrasound modality as a data augmentation technique applied to downstream classification tasks. This study aims to explore and evaluate the generation of synthetic ultrasound fetal brain images via GANs and apply them to improve fetal brain ultrasound plane classification. State of the art GANs stylegan2-ada were applied to fetal brain image generation and GAN-based data augmentation classifiers were compared with baseline classifiers. Our experimental results show that using data generated by both GANs and classical augmentation strategies allows for increasing the accuracy and area under the curve score.
| S-EPMC8659720 | biostudies-literature
Project description:Single-cell RNA sequencing (scRNA-Seq) is an emerging strategy for characterizing immune cell populations. Compared to flow or mass cytometry, scRNA-Seq could potentially identify cell types and activation states that lack precise cell surface markers. However, scRNA-Seq is currently limited due to the need to manually classify each immune cell from its transcriptional profile. While recently developed algorithms accurately annotate coarse cell types (e.g. T cells versus macrophages), making fine distinctions (e.g. CD8+ effector memory T cells) remains a difficult challenge. To address this, we developed a machine learning classifier called ImmClassifier that leverages a hierarchical ontology of cell type. We demonstrate that its predictions are highly concordant with flow-based markers from CITE-seq and outperforms other tools (+15% recall, +14% precision) in distinguishing fine-grained cell types with comparable performance on coarse ones. Thus, ImmClassifier can be used to explore more deeply the heterogeneity of the immune system in scRNA-Seq experiments.
| S-EPMC8536868 | biostudies-literature
Project description:Mathematical models of biochemical networks can largely facilitate the comprehension of the mechanisms at the basis of cellular processes, as well as the formulation of hypotheses that can be tested by means of targeted laboratory experiments. However, two issues might hamper the achievement of fruitful outcomes. On the one hand, detailed mechanistic models can involve hundreds or thousands of molecular species and their intermediate complexes, as well as hundreds or thousands of chemical reactions, a situation generally occurring in rule-based modeling. On the other hand, the computational analysis of a model typically requires the execution of a large number of simulations for its calibration, or to test the effect of perturbations. As a consequence, the computational capabilities of modern Central Processing Units can be easily overtaken, possibly making the modeling of biochemical networks a worthless or ineffective effort. To the aim of overcoming the limitations of the current state-of-the-art simulation approaches, we present in this paper FiCoS, a novel "black-box" deterministic simulator that effectively realizes both a fine-grained and a coarse-grained parallelization on Graphics Processing Units. In particular, FiCoS exploits two different integration methods, namely, the Dormand-Prince and the Radau IIA, to efficiently solve both non-stiff and stiff systems of coupled Ordinary Differential Equations. We tested the performance of FiCoS against different deterministic simulators, by considering models of increasing size and by running analyses with increasing computational demands. FiCoS was able to dramatically speedup the computations up to 855×, showing to be a promising solution for the simulation and analysis of large-scale models of complex biological processes.
| S-EPMC8476010 | biostudies-literature
Project description:The user alignment of cross-social networks is divided into user and group alignments, respectively. Obtaining users' full features is difficult due to social network privacy protection policies in user alignment mode. In contrast, the alignment accuracy is low due to the large number of edge users in the group alignment mode. To resolve this issue, First, stable topics are obtained from user-generated content (UGC) based on embedded topic jitter time, and the weight of user edges is updated by using vector distances. An improved Louvain algorithm, called Stable Topic-Louvain (ST-L), is designed to accomplish multi-level community detection without predetermined tags. It aims to obtain fuzzy topic features of the community and finalize the community alignment across social networks. Furthermore, iterative alignment is executed from coarse-grained communities to fine-grained sub-communities until user-level alignment occurs. The process can be terminated at any layer to achieve multi-granularity alignment, which resolves the low accuracy issue of edge user alignment at a single granularity and improves the accuracy of user alignment. The effectiveness of the proposed method is shown by implementing real datasets.
| S-EPMC10909205 | biostudies-literature
Project description:Long terminal repeat (LTR) retrotransposons and endogenous retroviruses (ERVs) are transposable elements in eukaryotic genomes well suited for computational identification. De novo identification tools determine the position of potential LTR retrotransposon or ERV insertions in genomic sequences. For further analysis, it is desirable to obtain an annotation of the internal structure of such candidates. This article presents LTRdigest, a novel software tool for automated annotation of internal features of putative LTR retrotransposons. It uses local alignment and hidden Markov model-based algorithms to detect retrotransposon-associated protein domains as well as primer binding sites and polypurine tracts. As an example, we used LTRdigest results to identify 88 (near) full-length ERVs in the chromosome 4 sequence of Mus musculus, separating them from truncated insertions and other repeats. Furthermore, we propose a work flow for the use of LTRdigest in de novo LTR retrotransposon classification and perform an exemplary de novo analysis on the Drosophila melanogaster genome as a proof of concept. Using a new method solely based on the annotations generated by LTRdigest, 518 potential LTR retrotransposons were automatically assigned to 62 candidate groups. Representative sequences from 41 of these 62 groups were matched to reference sequences with >80% global sequence similarity.
| S-EPMC2790888 | biostudies-literature
Project description:Genome-scale metabolic models (GEMs) are powerful tools for predicting cellular metabolic and physiological states. However, there are still missing reactions in GEMs due to incomplete knowledge. Recent gaps filling methods suggest directly predicting missing responses without relying on phenotypic data. However, they do not differentiate between substrates and products when constructing the prediction models, which affects the predictive performance of the models. In this paper, we propose a hyperedge prediction model that distinguishes substrates and products based on dual-scale fused hypergraph convolution, DSHCNet, for inferring the missing reactions to effectively fill gaps in the GEM. First, we model each hyperedge as a heterogeneous complete graph and then decompose it into three subgraphs at both homogeneous and heterogeneous scales. Then we design two graph convolution-based models to, respectively, extract features of the vertices in two scales, which are then fused via the attention mechanism. Finally, the features of all vertices are further pooled to generate the representative feature of the hyperedge. The strategy of graph decomposition in DSHCNet enables the vertices to engage in message passing independently at both scales, thereby enhancing the capability of information propagation and making the obtained product and substrate features more distinguishable. The experimental results show that the average recovery rate of missing reactions obtained by DSHCNet is at least 11.7% higher than that of the state-of-the-art methods, and that the gap-filled GEMs based on our DSHCNet model achieve the best prediction performance, demonstrating the superiority of our method.
| S-EPMC11299038 | biostudies-literature
Project description:Vegetation mapping is of considerable significance to both geoscience and mountain ecology, and the improved resolution of remote sensing images makes it possible to map vegetation at a finer scale. While the automatic classification of vegetation has gradually become a research hotspot, real-time and rapid collection of samples has become a bottleneck. How to achieve fine-scale classification and automatic sample selection at the same time needs further study. Stratified sampling based on appropriate prior knowledge is an effective sampling method for geospatial objects. Therefore, based on the idea of stratified sampling, this paper used the following three steps to realize the automatic selection of representative samples and classification of fine-scale mountain vegetation: 1) using Mountain Altitudinal Belt (MAB) distribution information to stratify the study area into multiple vegetation belts; 2) selecting and correcting samples through iterative clustering at each belt automatically; 3) using RF (Random Forest) classifier with strong robustness to achieve automatic classification. The average sample accuracy of nine vegetation formations was 0.933, and the total accuracy of the classification result was 92.2%, with the kappa coefficient of 0.910. The results showed that this method could automatically select high-quality samples and obtain a high-accuracy vegetation map. Compared with the traditional vegetation mapping method, this method greatly improved the efficiency, which is of great significance for the fine-scale mountain vegetation mapping in large-scale areas.
| S-EPMC7447069 | biostudies-literature
Project description:Alzheimer's disease (AD) may cause damage to the memory cells permanently, which results in the form of dementia. The diagnosis of Alzheimer's disease at an early stage is a problematic task for researchers. For this, machine learning and deep convolutional neural network (CNN) based approaches are readily available to solve various problems related to brain image data analysis. In clinical research, magnetic resonance imaging (MRI) is used to diagnose AD. For accurate classification of dementia stages, we need highly discriminative features obtained from MRI images. Recently advanced deep CNN-based models successfully proved their accuracy. However, due to a smaller number of image samples available in the datasets, there exist problems of over-fitting hindering the performance of deep learning approaches. In this research, we developed a Siamese convolutional neural network (SCNN) model inspired by VGG-16 (also called Oxford Net) to classify dementia stages. In our approach, we extend the insufficient and imbalanced data by using augmentation approaches. Experiments are performed on a publicly available dataset open access series of imaging studies (OASIS), by using the proposed approach, an excellent test accuracy of 99.05% is achieved for the classification of dementia stages. We compared our model with the state-of-the-art models and discovered that the proposed model outperformed the state-of-the-art models in terms of performance, efficiency, and accuracy.
| S-EPMC7071616 | biostudies-literature
Project description:In the Information and Communication Technology age, connected objects generate massive amounts of data traffic, which enables data analysis to uncover previously hidden trends and detect unusual network-load. We identify five core design principles to consider when designing a deep learning-empowered intrusion detection system (IDS). We proposed the Temporal Convolution Neural Network (TCNN), an intelligent model for IoT-IDS that aggregates convolution neural network (CNN) and generic convolution, based on these concepts. To handle unbalanced datasets, TCNN is accumulated with synthetic minority oversampling technique with nominal continuity. It is also used in conjunction with effective feature engineering techniques like attribute transformation and reduction. The presented model is compared to two traditional machine learning algorithms, random forest (RF) and logistic regression (LR), as well as LSTM and CNN deep learning techniques, using the Bot-IoT data repository. The outcomes of the experiments depicts that TCNN maintains a strong balance of efficacy and performance. It is better as compared to other deep learning IDSs, with a multi-class traffic detection accuracy of 99.9986 percent and a training period that is very close to CNN.
| S-EPMC8507481 | biostudies-literature