Non-deep Learning Research Articles

Global Analysis of Deep Learning Prediction Using Large-Scale In-House Kinome-Wide Profiling Data.

In drug discovery, the prediction of activity and absorption, distribution, metabolism, excretion, and toxicity parameters is one of the most important approaches in determining which compound to synthesize next. In recent years, prediction methods based on deep learning as well as non-deep learning approaches have been established, and a number of applications to drug discovery have been reported by various companies and organizations. In this research, we performed activity prediction using deep learning and non-deep learning methods on in-house assay data for several hundred kinases and compared and discussed the prediction results. We found that the prediction accuracy of the single-task graph neural network (GNN) model was generally lower than that of the non-deep learning model (LightGBM), but the multitask GNN model, which combined data from other kinases, comprehensively outperformed LightGBM. In addition, the extrapolative validity of the multitask model was verified by using it for prediction on known kinase ligands. We observed an overlap between characteristic protein–ligand interaction sites and the atoms that are important for prediction. By building appropriate models based on the conditions of the data set and analyzing the feature importance of the prediction results, a ligand-based prediction method may be used not only for activity prediction but also for drug design.

Unsupervised Deep Learning Registration of Uterine Cervix Sequence Images.

Simple SummaryA sequence of images can be taken after the application of acetic acid during a colposcopic examination of the uterine cervix to capture the dynamic visual variations due to the aceto-whitening reaction on the cervical epithelium. Automated analyis of these changes require spatial alignment of the apparent change in the cervix location in the image sequence due to patient movement or imaging device repositioning. We developed a new self-supervised RGB-colored deep learning-based image registration method to automatically align the images that does not require a manually-provided reference standard. We also fine-tuned a transformer-based segmentation network to evaluate the result of our registration method which achieved 12.62% higher in Dice/IoU scores in cervix boundary detection compared to the unregistered images.During a colposcopic examination of the uterine cervix for cervical cancer prevention, one or more digital images are typically acquired after the application of diluted acetic acid. An alternative approach is to acquire a sequence of images at fixed intervals during an examination before and after applying acetic acid. This approach is asserted to be more informative as it can capture dynamic pixel intensity variations on the cervical epithelium during the aceto-whitening reaction. However, the resulting time sequence images may not be spatially aligned due to the movement of the cervix with respect to the imaging device. Disease prediction using automated visual evaluation (AVE) techniques using multiple images could be adversely impacted without correction for this misalignment. The challenge is that there is no registration ground truth to help train a supervised-learning-based image registration algorithm. We present a novel unsupervised registration approach to align a sequence of digital cervix color images. The proposed deep-learning-based registration network consists of three branches and processes the red, green, and blue (RGB, respectively) channels of each input color image separately using an unsupervised strategy. Each network branch consists of a convolutional neural network (CNN) unit and a spatial transform unit. To evaluate the registration performance on a dataset that has no ground truth, we propose an evaluation strategy that is based on comparing automatic cervix segmentation masks in the registered sequence and the original sequence. The compared segmentation masks are generated by a fine-tuned transformer-based object detection model (DeTr). The segmentation model achieved Dice/IoU scores of 0.917/0.870 and 0.938/0.885, which are comparable to the performance of our previous model in two datasets. By comparing our segmentation on both original and registered time sequence images, we observed an average improvement in Dice scores of 12.62% following registration. Further, our approach achieved higher Dice and IoU scores and maintained full image integrity compared to a non-deep learning registration method on the same dataset.

CryoSat-2 waveform classification for melt event monitoring

Measuring the mass balance of ice sheets is important with respect to understanding among others sea level rise, glacier dynamics, global ocean circulation and marine ecosystems. One important parameter of the mass balance is surface melt, which can be estimated from different satellite data sources. In this study we investigate the potential of utilizing machine learning techniques for CryoSat-2 (CS2) radar altimeter waveform classification in order to derive melt information. Training data is derived by spatio-temporally matching of CS2 measurements with MODIS land surface temperature measurements. We propose a time convolution network with a fully connected classifier tail for CS2 waveform classifcation. In addition a non-deep learning model is implemented, providing a baseline. One of the main challenges is the high class imbalance, as surface temperatures on the interior of Greenland rarely reach the freezing point. The model performance is measured by several metrics: F1 score, average recall and Matthews correlation coefficient. The results of this proof of concept study indicate feasibility.

Deep Learning-Based 3D Shape Feature Extraction on Flash Animation Style

Flash animation, as a kind of digital learning resource, is an important media for delivering information content, and more importantly, it is an important online learning resource with text, graphics, images, audio, video, interaction, dynamic effects, etc. Flash animation, with its powerful multimedia interaction and presentation capabilities, is widely used in distance education, high-quality course websites, Q&A platforms, etc. With the continuous development of deep learning, the 3D shape feature extraction method combined with deep learning has become a hot research topic. In this paper, we combine deep learning with traditional 3D shape feature extraction methods, so that we can not only break the bottleneck of nondeep learning methods but also improve the accuracy of 3D shape data classification and retrieval tasks, especially in the case of non-rigid 3D shapes. The scheme in this paper not only does not require a large number of training samples but also its feature extraction for flash animation is accurate. Experiments show that the success rate of accurate feature extraction of this paper’s scheme is higher than that of the state-of-the-art methods.

Recognition of mRNA N4 Acetylcytidine (ac4C) by Using Non-Deep vs. Deep Learning

Deep learning models have been successfully applied in a wide range of fields. The creation of a deep learning framework for analyzing high-performance sequence data have piqued the research community’s interest. N4 acetylcytidine (ac4C) is a post-transcriptional modification in mRNA, is an mRNA component that plays an important role in mRNA stability control and translation. The ac4C method of mRNA changes is still not simple, time consuming, or cost effective for conventional laboratory experiments. As a result, we developed DL-ac4C, a CNN-based deep learning model for ac4C recognition. In the alternative scenario, the model families are well-suited to working in large datasets with a large number of available samples, especially in biological domains. In this study, the DL-ac4C method (deep learning) is compared to non-deep learning (machine learning) methods, regression, and support vector machine. The results show that DL-ac4C is more advanced than previously used approaches. The proposed model improves the accuracy recall area by 9.6 percent and 9.8 percent, respectively, for cross-validation and independent tests. More nuanced methods of incorporating prior bio-logical knowledge into the estimation procedure of deep learning models are required to achieve better results in terms of predictive efficiency and cost-effectiveness. Based on an experiment’s acetylated dataset, the DL-ac4C sequence-based predictor for acetylation sites in mRNA can predict whether query sequences have potential acetylation motifs.

Robust face recognition for occluded real‐world images using constrained probabilistic sparse network

Aiming at the occluded real-world face images across illumination, pose, expression, and resolution variations, a robust face recognition for occluded real-world images using constrained probabilistic sparse network is presented. A constrained probabilistic sparse representation network is constructed to obtain the features of all the training images from a global perspective, and the new network nodes are generated through the random combination of the training images. In the probabilistic sparse representation network, the probabilities of each class of the sparse subspace that the occluded test images individually belong to are defined and calculated. The final classifications of the test images are determined by the joint maximum probability of the network nodes. Meanwhile the second-order gradient constraint is the first introduced in the probabilistic sparse representation network. It is found that the constraint uses the adjacent pixels of the face images to obtain the local texture similarity, and further use the local texture similarity to distinguish the occlusion and non-occlusion parts. Thus the constraint can reduce the influence of the occlusion part on face recognition. Extensive experiments with the 12 existing methods on the five face databases demonstrate that the recognition rate of the proposed method is the best than the non-deep learning methods compared, and the proposed method can obtain nearly the same recognition rate with an advantage of a very less time consumption compared to the state-of-the-art deep learning methods.

Automated 3D U‐net based segmentation of neonatal cerebral ventricles from 3D ultrasound images

Intraventricular hemorrhaging (IVH) within cerebral lateral ventricles affects 20-30% of very low birth weight infants (<1500g). As the ventricles increase in size, the intracranial pressure increases, leading to post-hemorrhagic ventricle dilatation (PHVD), an abnormal enlargement of the head. The most widely used imaging tool for measuring IVH and PHVD is cranial two-dimensional (2D) ultrasound (US). Estimating volumetric changes over time with 2D US is unreliable due to high user variability when locating the same anatomical location at different scanning sessions. Compared to 2D US, three-dimensional (3D) US is more sensitive to volumetric changes in the ventricles and does not suffer from variability in slice acquisition. However, 3D US images require segmentation of the ventricular surface, which is tedious and time-consuming when done manually. A fast, automated ventricle segmentation method for 3D US would provide quantitative information in a timely manner when monitoring IVH and PHVD in pre-term neonates. To this end, we developed a fast and fully automated segmentation method to segment neonatal cerebral lateral ventricles from 3D US images using deep learning. Our method consists of a 3D U-Net ensemble model composed of three U-Net variants, each highlighting various aspects of the segmentation task such as the shape and boundary of the ventricles. The ensemble is made of a U-Net++, attention U-Net, and U-Net with a deep learning-based shape prior combined using a mean voting strategy. We used a dataset consisting of 190 3D US images, which was separated into two subsets, one set of 87 images contained both ventricles, and one set of 103 images contained only one ventricle (caused by limited field-of-view during acquisition). We conducted fivefold cross-validation to evaluate the performance of the models on a larger amount of test data; 165 test images of which 75 have two ventricles (two-ventricle images) and 90 have one ventricle (one-ventricle images). We compared these results to each stand-alone model and to previous works including, 2D multiplane U-Net and 2D SegNet models. Using fivefold cross-validation, the ensemble method reported a Dice similarity coefficient (DSC) of 0.720 ± 0.074, absolute volumetric difference (VD) of 3.7 ± 4.1 cm3 , and a mean absolute surface distance (MAD) of 1.14 ± 0.41mm on 75 two-ventricle test images. Using 90 test images with a single ventricle, the model after cross-validation reported DSC, VD, and MAD values of 0.806 ± 0.111, 3.5 ± 2.9 cm3 , and 1.37 ± 1.70mm, respectively. Compared to alternatives, the proposed ensemble yielded a higher accuracy in segmentation on both test data sets. Our method required approximately 5 s to segment one image and was substantially faster than the state-of-the-art conventional methods. Compared to the state-of-the-art non-deep learning methods, our method based on deep learning was more efficient in segmenting neonatal cerebral lateral ventricles from 3D US images with comparable or better DSC, VD, and MAD performance. Our dataset was the largest to date (190 images) for this segmentation problem and the first to segment images that show only one lateral cerebral ventricle.

Deep Learning Accurately Quantifies Plasma Cell Percentages on CD138-Stained Bone Marrow Samples

The diagnosis of plasma cell neoplasms requires accurate, and ideally precise, percentages. This plasma cell percentage is often determined by visual estimation of CD138-stained bone marrow biopsies and clot sections. While not necessarily inaccurate, estimates are by definition imprecise. For this study, we hypothesized that deep learning can be used to improve precision. We trained a semantic segmentation-based convolutional neural network (CNN) using annotations of CD138+ and CD138- cells provided by one pathologist on small image patches of bone marrow and validated the CNN on an independent test set of image patches using annotations from two pathologists and a non-deep learning commercial software. On validation, we found that the intraclass correlation coefficients for plasma cell percentages between the CNN and pathologist #1, a non-deep learning commercial software and pathologist #1, and pathologists #1 and #2 were 0.975, 0.892, and 0.994, respectively. The overall results show that CNN labels were almost as accurate as pathologist labels at a cell-by-cell level. Once satisfied with performance, we scaled-up the CNN to evaluate whole slide images (WSIs), and deployed the system as a workflow friendly web application to measure plasma cell percentages using snapshots taken from microscope cameras.

Pay Attention to Local Contrast Learning Networks for Infrared Small Target Detection

Infrared small target suffers from the lack of intrinsic features, context and samples. Conventional detection methods are usually unable to sufficiently and effectively extract the features of infrared small targets. Therefore, we propose a novel attention-based local contrast learning network (ALCL-Net). Considering the scarcity of intrinsic features of infrared small targets, we propose ResNet32, which enhances the ability to extract infrared small target features and avoids the problem that the target features are overwhelmed by the background features due to too deep network. At the same time, we construct a simplified bilinear interpolation attention module (SBAM), which is used for fusion of hierarchical feature maps. It has fast inference speed and can focus on the feature of the target in the lack of context. Furthermore, local contrast learning (LCL) is introduced, which adopts the local contrast idea of non-deep learning methods. It can alleviate the dependence on dataset samples, thereby improving detection accuracy on datasets with few samples. Compared with the state-of-the-art methods, the proposed ALCL-Net achieves superior performance with an intersection-over-union (IoU) of 0.792 and normalized IoU (nIoU) of 0.771 on the public SIRST dataset.

Stochastic Learning-Based Artificial Neural Network Model for an Automatic Tuberculosis Detection System Using Chest X-Ray Images

Tuberculosis (TB) is still one of the most serious health issues today with a high fatality rate. While attempts are being made to make primary diagnosis more reliable and accessible in places with high tuberculosis rates, Chest X-rays has become a popular source. However, specialist radiologists are required for the screening process, which could be a challenge in developing countries. For early diagnosis of tuberculosis utilizing CXR images, a complete automatic system of tuberculosis detection can decrease the need for trained staff. Various deep learning and machine learning technologies have been introduced in recent years for examining digital chest radiographs for TB-related variances with the goal of reducing inter-class reader variability and reproducibility, as well as providing radiologic services in areas where radiologists are not available. Tuberculosis is sometimes misclassified as other conditions with similar radiographic patterns as a result of CXR images, resulting in inefficient therapy. The current approach, however, is limited to Computer-Aided Detection (CAD), which has only been evaluated with non-deep learning models. Deep neural networks open potentially new avenues for tuberculosis treatment. There are no peer-reviewed studies comparing the effectiveness of various deep learning systems in detecting TB anomalies, and none compare multiple deep learning systems with human readers. In this paper, the aim of the proposed method is to develop an efficient tuberculosis detection system based on stochastic learning with artificial neural network (ANN) model by random variations using Chest X-ray images. This approach can able to incorporate random functions into the network, either by assigning stochastic transfer functions to the network or by assigning stochastic weights to the network. This proposed method is to learn features from CXR images and optimize the parameters of an ANN model by randomly mixing the training dataset before each iteration, resulting in varied ordering of model parameter updates. Furthermore, in a neural network, model weights are frequently initialized at a random beginning point. By focusing on randomness functions with optimization, the proposed technique achieved great accuracy. The motivation of the proposed method is to detect abnormalities in CXR with the different levels of complexity of TB by strong or weak evidence with different deep geometric contexts such as shape, size, cavitation, and density. ANN’s primary benefit is extracting hidden linear and non-linear relationships of high-dimensional and complex data. The proposed method was thoroughly tested with the Shenzhen and Montgomery datasets using metrics such as sensitivity, specificity, and accuracy, and it was discovered that the proposed method attained better accuracy when compared to state-of-the-art methods. The proposed method shows an improved efficiency with sensitivity of 96.12%, specificity of 98.01%, accuracy 98.45% and F-Score 95.88% respectively.

Artificial Intelligence: Innovation to Assist in the Identification of Sono-anatomy for Ultrasound-Guided Regional Anaesthesia.

Ultrasound-guided regional anaesthesia (UGRA) involves the targeted deposition of local anaesthesia to inhibit the function of peripheral nerves. Ultrasound allows the visualisation of nerves and the surrounding structures, to guide needle insertion toa perineural or fascial plane end point for injection. However, it is challenging to develop the necessary skills to acquire and interpret optimal ultrasound images. Sound anatomical knowledge is required and human image analysis is fallible, limited by heuristic behaviours and fatigue, while its subjectivity leads to varied interpretation even amongst experts. Therefore, to maximise the potential benefit of ultrasound guidance, innovation in sono-anatomical identification is required.Artificial intelligence (AI) is rapidly infiltrating many aspects of everyday life. Advances related to medicine have been slower, in part because of the regulatory approval process needing to thoroughly evaluate the risk-benefit ratio of new devices. One area of AI to show significant promise is computer vision (a branch of AI dealing with how computers interpret the visual world), which is particularly relevant to medical image interpretation. AI includes the subfields of machine learning and deep learning, techniques used to interpret or label images. Deep learning systems may hold potential to support ultrasound image interpretation in UGRA but must be trained and validated on data prior to clinical use.Review of the current UGRA literature compares the success and generalisability of deep learning and non-deep learning approaches to image segmentation and explains how computers are able to track structures such as nerves through image frames. We conclude this review with a case study from industry (ScanNav Anatomy Peripheral Nerve Block; Intelligent Ultrasound Limited). This includes a more detailed discussion of the AI approach involved in this system and reviews current evidence of the system performance.The authors discuss how this technology may be best used to assist anaesthetists and what effects this may have on the future of learning and practice of UGRA. Finally, we discuss possible avenues for AI within UGRA and the associated implications.

Automatic stomata recognition and measurement based on improved YOLO deep learning model and entropy rate superpixel algorithm

The traditional methods of analyzing stomatal traits are mostly manual observation and measurement. These methods are time-consuming, labor-intensive, and inefficient. Some methods have been proposed for the automatic recognition and counting of stomata, however most of those methods could not complete the automatic measurement of stomata parameters at the same time. Some non-deep learning methods could automatically measure the parameters of stomata, but they could not complete the automatic recognition and detection of stomata. In this paper, a deep learning-based method was proposed for automatically identifying, counting and measuring stomata of maize (Zea mays L.) leaves at the same time. An improved YOLO (You Only Look Once) deep learning model was proposed to identify stomata of maize leaves automatically, and an entropy rate superpixel algorithm was used for the accurate measurement of stomatal parameters. According to the characteristics of the stomata images data set, the network structure of YOLOv5 was modified, which greatly reduced the training time without affecting the recognition performance. The predictor in YOLO deep learning model was optimized, which reduced the false detection rate. At the same time, the 16-fold and 32-fold down-sampling layers were simplified according to the characteristics of stomatal objects, which improved the recognition efficiency. Experimental results showed that the recognition precision of the improved YOLO deep learning model reached 95.3% on the maize leaves stomatal data set, and the average accuracy of parameter measurement reached 90%. The proposed method could fully automatically complete the recognition, counting and measurement of stomata of plants, which can help agricultural scientists and botanists to conduct large-scale researches of stomatal morphology, structure and physiology, as well as the researches combined with genetic analysis or molecular-level analysis.

Adaptive Decision Support System for On-Line Multi-Class Learning and Object Detection

A new online multi-class learning algorithm is proposed with three main characteristics. First, in order to make the feature pool fitter for the pattern pool, the adaptive feature pool is proposed to dynamically combine the three general features, Haar-like, Histogram of Oriented Gradient (HOG), and Local Binary Patterns (LBP). Second, the external model is integrated into the proposed model without re-training to enhance the efficacy of the model. Third, a new multi-class learning and updating mechanism are proposed that help to find unsuitable decisions and adjust them automatically. The performance of the proposed model is validated with multi-class detection and online learning system. The proposed model achieves a better score than other non-deep learning algorithms used in public pedestrian and multi-class databases. The multi-class databases contain data for pedestrians, faces, vehicles, motorcycles, bicycles, and aircraft.

Prediction of Merchandise Sales on E-Commerce Platforms Based on Data Mining and Deep Learning

Online business has grown exponentially during the last decade, and the industries are focusing on online business more than before. However, just setting up an online store and starting selling might not work. Different machine learning and data mining techniques are needed to know the users’ preferences and know what would be best for business. According to the decision-making needs of online product sales, combined with the influencing factors of online product sales in various industries and the advantages of deep learning algorithm, this paper constructs a sales prediction model suitable for online products and focuses on evaluating the adaptability of the model in different types of online products. In the research process, the full connection model is compared with the training results of CNN, which proves the accuracy and generalization ability of CNN model. By selecting the non-deep learning model as the comparison baseline, the performance advantages of CNN model under different categories of products are proved. In addition, the experiment concludes that the unsupervised pretrained CNN model is more effective and adaptable in sales forecasting.

SCANet: A Spatial and Channel Attention based Network for Partial-to-Partial Point Cloud Registration

Point cloud registration plays an essential role in many areas, such as computer vision and robotics. However, traditional feature-based registration requires handcrafted descriptors for various scenarios, which is of low efficiency and flexibility; ICP and its locally optimal variants are sensitive to initialization, while globally optimal methods are of high computational time to overcome noise, outliers, and partial overlap. Learning-based registration can automatically and flexibly learn shape representation for different objects, but existing methods are of either low efficiency or low precision, and poorly perform in partial-to-partial point cloud registration. Thus, we present a simple spatial and channel attention based network, named SCANet, for partial-to-partial point cloud registration. A spatial self-attention aggregation (SSA) module is applied in a feature extraction sub-network to efficiently make use of the inter and global information of each point cloud in different levels, while a channel cross-attention regression (CCR) module is adopted in a pose estimation sub-network for information interaction between two input global feature vectors, enhancing relevant information and suppressing redundant information. Experimental results show that our SCANet achieves state-of-the-art performances in both accuracy and efficiency compared to existing non-deep learning and learning-based methods on partial visibility with Gaussian noise. Our source code is available at the project website https://github.com/zhouruqin/SCANet.

Deep/Transfer Learning with Feature Space Ensemble Networks (FeatSpaceEnsNets) and Average Ensemble Networks (AvgEnsNets) for Change Detection Using DInSAR Sentinel-1 and Optical Sentinel-2 Satellite Data Fusion

Differential interferometric synthetic aperture radar (DInSAR), coherence, phase, and displacement are derived from processing SAR images to monitor geological phenomena and urban change. Previously, Sentinel-1 SAR data combined with Sentinel-2 optical imagery has improved classification accuracy in various domains. However, the fusing of Sentinel-1 DInSAR processed imagery with Sentinel-2 optical imagery has not been thoroughly investigated. Thus, we explored this fusion in urban change detection by creating a verified balanced binary classification dataset comprising 1440 blobs. Machine learning models using feature descriptors and non-deep learning classifiers, including a two-layer convolutional neural network (ConvNet2), were used as baselines. Transfer learning by feature extraction (TLFE) using various pre-trained models, deep learning from random initialization, and transfer learning by fine-tuning (TLFT) were all evaluated. We introduce a feature space ensemble family (FeatSpaceEnsNet), an average ensemble family (AvgEnsNet), and a hybrid ensemble family (HybridEnsNet) of TLFE neural networks. The FeatSpaceEnsNets combine TLFE features directly in the feature space using logistic regression. AvgEnsNets combine TLFEs at the decision level by aggregation. HybridEnsNets are a combination of FeatSpaceEnsNets and AvgEnsNets. Several FeatSpaceEnsNets, AvgEnsNets, and HybridEnsNets, comprising a heterogeneous mixture of different depth and architecture models, are defined and evaluated. We show that, in general, TLFE outperforms both TLFT and classic deep learning for the small dataset used and that larger ensembles of TLFE models do not always improve accuracy. The best performing ensemble is an AvgEnsNet (84.862%) comprised of a ResNet50, ResNeXt50, and EfficientNet B4. This was matched by a similarly composed FeatSpaceEnsNet with an F1 score of 0.001 and variance of 0.266 less. The best performing HybridEnsNet had an accuracy of 84.775%. All of the ensembles evaluated outperform the best performing single model, ResNet50 with TLFE (83.751%), except for AvgEnsNet 3, AvgEnsNet 6, and FeatSpaceEnsNet 5. Five of the seven similarly composed FeatSpaceEnsNets outperform the corresponding AvgEnsNet.

Self-supervised pretraining for transferable quantitative phase image cell segmentation.

In this paper, a novel U-Net-based method for robust adherent cell segmentation for quantitative phase microscopy image is designed and optimised. We designed and evaluated four specific post-processing pipelines. To increase the transferability to different cell types, non-deep learning transfer with adjustable parameters is used in the post-processing step. Additionally, we proposed a self-supervised pretraining technique using nonlabelled data, which is trained to reconstruct multiple image distortions and improved the segmentation performance from 0.67 to 0.70 of object-wise intersection over union. Moreover, we publish a new dataset of manually labelled images suitable for this task together with the unlabelled data for self-supervised pretraining.

Detecting Salient Image Objects Using Color Histogram Clustering for Region Granularity.

Salient object detection represents a novel preprocessing stage of many practical image applications in the discipline of computer vision. Saliency detection is generally a complex process to copycat the human vision system in the processing of color images. It is a convoluted process because of the existence of countless properties inherent in color images that can hamper performance. Due to diversified color image properties, a method that is appropriate for one category of images may not necessarily be suitable for others. The selection of image abstraction is a decisive preprocessing step in saliency computation and region-based image abstraction has become popular because of its computational efficiency and robustness. However, the performances of the existing region-based salient object detection methods are extremely hooked on the selection of an optimal region granularity. The incorrect selection of region granularity is potentially prone to under- or over-segmentation of color images, which can lead to a non-uniform highlighting of salient objects. In this study, the method of color histogram clustering was utilized to automatically determine suitable homogenous regions in an image. Region saliency score was computed as a function of color contrast, contrast ratio, spatial feature, and center prior. Morphological operations were ultimately performed to eliminate the undesirable artifacts that may be present at the saliency detection stage. Thus, we have introduced a novel, simple, robust, and computationally efficient color histogram clustering method that agglutinates color contrast, contrast ratio, spatial feature, and center prior for detecting salient objects in color images. Experimental validation with different categories of images selected from eight benchmarked corpora has indicated that the proposed method outperforms 30 bottom-up non-deep learning and seven top-down deep learning salient object detection methods based on the standard performance metrics.

Time series remote sensing image classification framework using combination of deep learning and multiple classifiers system

Recently, time series image (TSI) has been reported to be an effective resource to mapping fine land use/land cover (LULC), and deep learning, in particular, has been gaining growing attention in this field. However, deep learning methods using single classifier need further improvement for accurate TSI classification owing to the 1D temporal properties and insufficient dense time series of the remote sensing images. To overcome such disadvantages, we proposed an innovative approach involving construction of TSI and combination of deep learning and multiple classifiers system (MCS). Firstly, we used a normalised difference index (NDI) to establish an NDIs-based TSI and then designed a framework consisting of a deep learning-based feature extractor and multiple classifiers system (MCS) based classification model to classify the TSI. With the new approach, our experiments were conducted on Landsat images located in two counties, Sutter and Kings in California, United States. The experimental results indicate that our proposed method achieves great progress on accuracy improvement and LULC mapping, outperforming classifications using comparative deep learning and non-deep learning methods.

Exploring dropout discriminator for domain adaptation

Adaptation of a classifier to new domains is one of the challenging problems in machine learning. This has been addressed using many deep and non-deep learning based methods. Among the methodologies used, that of adversarial learning is widely applied to solve many deep learning problems along with domain adaptation. These methods are based on a discriminator that ensures source and target distributions are close. However, here we suggest that rather than using a point estimate obtaining by a single discriminator, it would be useful if a distribution based on ensembles of discriminators could be used to bridge this gap. This could be achieved using multiple classifiers or using traditional ensemble methods. In contrast, we suggest that a Monte Carlo dropout based ensemble discriminator could suffice to obtain the distribution based discriminator. Specifically, we propose a curriculum based dropout discriminator that gradually increases the variance of the sample based distribution and the corresponding reverse gradients are used to align the source and target feature representations. An ensemble of discriminators helps the model to learn the data distribution efficiently. It also provides a better gradient estimates to train the feature extractor. The detailed results and thorough ablation analysis show that our model outperforms state-of-the-art results.