In this paper, we present a patch-based land use and cover classification approach using Sentinel-2 satellite images. The images are openly freely accessible, provided in the earth observation program Copernicus. We novel dataset, based on these that covers 13 spectral bands is comprised of ten classes with total 27 000 labeled geo-referenced Benchmarks for dataset its state-of-the-art deep convolutional neural networks. An overall accuracy 98.57% was achieved proposed dataset. resulting...

Traditional distance and density-based anomaly detection techniques are unable to detect periodic seasonality related point anomalies which occur commonly in streaming data, leaving a big gap time series the current era of IoT. To address this problem, we present novel deep learning-based approach (DeepAnT) for is equally applicable non-streaming cases. DeepAnT capable detecting wide range anomalies, i.e., contextual discords data. In contrast methods where learned, uses unlabeled data...

Recognition of text in natural scene images is becoming a prominent research area due to the widespread availablity imaging devices low-cost consumer products like mobile phones. To evaluate performance recent algorithms detecting and recognizing from complex images, ICDAR 2011 Robust Reading Competition was organized. Challenge 2 competition dealt specifically with detecting/recognizing images. This paper presents an overview approaches that participants used, evaluation measure, dataset...

This paper presents a novel end-to-end system for table understanding in document images called DeepDeSRT. In particular, the contribution of DeepDeSRT is two-fold. First, it deep learning-based solution detection images. Secondly, proposes approach structure recognition, i.e. identifying rows, columns, and cell positions detected tables. contrast to existing rule-based methods, which rely on heuristics or additional PDF metadata (like, example, print instructions, character bounding boxes,...

The increased availability of high-resolution satellite imagery allows to sense very detailed structures on the surface our planet. Access such information opens up new directions in analysis remote sensing imagery. While deep neural networks have achieved significant advances semantic segmentation images, most existing approaches tend produce predictions with poor boundaries. In this paper, we address problem preserving boundaries by introducing a novel multi-task loss. loss leverages...

Light microscopy combined with well-established protocols of two-dimensional cell culture facilitates high-throughput quantitative imaging to study biological phenomena. Accurate segmentation individual cells in images enables exploration complex questions, but can require sophisticated processing pipelines cases low contrast and high object density. Deep learning-based methods are considered state-of-the-art for image typically vast amounts annotated data, which there is no suitable...

In this paper, we address the challenge of land use and cover classification using Sentinel-2 satellite images. The key contributions are as follows. We present a novel dataset based on images covering 13 different spectral bands consisting 10 classes with in total 27,000 labeled evaluate state-of-the-art deep Convolutional Neural Networks (CNNs) its bands. also CNNs existing remote sensing datasets compare obtained results. With proposed dataset, achieved an overall accuracy 98.57%. system...

The rapidly evolving field of sound classification has greatly benefited from the methods other domains. Today, trend is to fuse domain-specific tasks and approaches together, which provides community with new outstanding models.We present AudioCLIP – an extension CLIP model that handles audio in addition text images. Utilizing AudioSet dataset, our proposed incorporates ESResNeXt audio-model into framework, thus enabling it perform multimodal keeping CLIP's zero-shot capabilities.AudioCLIP...

With the advancement of powerful image processing and machine learning techniques, CAD has become ever more prevalent in all fields medicine including ophthalmology. Since optic disc is most important part retinal fundus for glaucoma detection, this paper proposes a two-stage framework that first detects localizes then classifies it into healthy or glaucomatous. The stage based on RCNN responsible localizing extracting from while second uses Deep CNN to classify extracted In addition...

Abstract We present Segment Anything for Microscopy, a tool interactive and automatic segmentation tracking of objects in multi-dimensional microscopy data. Our method is based on Anything, vision foundation model image segmentation. extend it by training specialized models data that significantly improve quality wide range imaging conditions. also implement annotation tools (volumetric) tracking, speed up compared to established tools. work constitutes the first application microscopy,...

We demonstrate how information about eye blink frequency and head motion patterns derived from Google Glass sensors can be used to distinguish different types of high level activities. While it is well known that correlated with user activity, our aim show (1) data an unobtrusive, commercial platform which not a dedicated tracker good enough useful (2) adding significantly improves the recognition rates. The method evaluated on set experiment containing five activity classes (reading,...

This paper presents a novel approach for the detection of tables present in documents, leveraging potential deep neural networks. Conventional approaches table rely on heuristics that are error prone and specific to dataset. In contrast, presented harvests data recognize arbitrary layout. Most prior only applicable PDFs, whereas, directly works images making it generally any format. The is based combination deformable CNN with faster R-CNN/FPN. has fixed receptive field which problematic...

This paper presents a deep Convolutional Neural Network (CNN) based approach for document image classification. One of the main requirement CNN architecture is that they need huge number samples training. To overcome this problem we adopt which trained using big dataset containing millions i.e., ImageNet. The proposed work outperforms both traditional structure similarity methods and approaches earlier. accuracy with merely 20 images per class state-of-the-art by achieving classification...

Propensity of skin diseases to manifest in a variety forms, lack and maldistribution qualified dermatologists, exigency timely accurate diagnosis call for automated Computer-Aided Diagnosis (CAD). This study aims at extending previous works on CAD dermatology by exploring the potential Deep Learning classify hundreds diseases, improving classification performance, utilizing disease taxonomy. We trained state-of-the-art Neural Networks two largest publicly available image datasets, namely...

In this contemporaneous world, it is an obligation for any organization working with documents to end up the insipid task of classifying truckload documents, which nascent stage venturing into realm information retrieval and data mining. But classification such humongous multiple classes, calls a lot time labor. Hence system could classify these acceptable accuracy would be unfathomable help in document engineering. We have created classifiers compared their on raw processed data. garnered...

This paper presents a novel framework for the demystification of convolutional deep learning models time-series analysis. is step toward making informed/explainable decisions in domain time series, powered by learning. There have been numerous efforts to increase interpretability image-centric neural network models, where learned features are more intuitive visualize. Visualization series significantly challenging, as there no direct interpretation filters and inputs compared with imaging...

Time series forecasting is one of the challenging problems for humankind. Traditional methods using mean regression models have severe shortcomings in reflecting real-world fluctuations. While new probabilistic rush to rescue, they fight with technical difficulties like quantile crossing or selecting a prior distribution. To meld different strengths these fields while avoiding their weaknesses as well push boundary state-of-the-art, we introduce ForGAN - step ahead generative adversarial...

The need for robust unsupervised anomaly detection in streaming data is increasing rapidly the current era of smart devices, where enormous are gathered from numerous sensors. These sensors record internal state a machine, external environment, and interaction machines with other humans. It prime importance to leverage this information order minimize downtime machines, or even avoid completely by constant monitoring. Since each device generates different type data, it normally case that...

Environmental Sound Classification (ESC) is an active research area in the audio domain and has seen a lot of progress past years. However, many existing approaches achieve high accuracy by relying on domain-specific features architectures, making it harder to benefit from advances other fields (e.g., image domain). Additionally, some successes have been attributed discrepancy how results are evaluated (i.e., unofficial splits UrbanSound8K (US8K) dataset), distorting overall progression...

Mental states like stress, depression, and anxiety have become a huge problem in our modern society. The main objective of this work is to detect stress among people, using Machine Learning approaches with the final aim improving their quality life. We propose various models for detection on individuals publicly available multimodal dataset, WESAD. Sensor data including electrocardiogram (ECG), body temperature (TEMP), respiration (RESP), electromyogram (EMG), electrodermal activity (EDA)...