Improvement In Overall Accuracy Research Articles

Fusion of Sentinel-1 and Sentinel-2 data in mapping the impervious surfaces at city scale.

Urbanization creates new development in open spaces and agricultural fields, synonymous with increasing impervious surfaces. Such surfaces restrain the natural infiltration of water, and directly affect the non-point source pollution. Thus, consequential events like flooding and surface water degradation require spatial and quantitative information on impervious surfaces. Remote sensing technologies are widely used in impervious surface mapping of various geographical locations for environmental monitoring. In this study, the datasets from recently launched European Space Agency satellites (Sentinel-1 and Sentinel-2) and random forest classifier are used. The impervious surface growth of the study area, Lahore city, in 2015 and 2021, and growth trends are assessed. Results are validated with classification accuracy and comparison with similar datasets. The objective is to develop a reliable impervious surface mapping method with land cover quantification technique from multisource datasets. With a chi-square value of greater than 3.84 obtained from the McNemar test, the performance of fused data was superior to that of optical alone data in the classification. Over a 5-year period, Lahore grew at an annual rate of 2.14% comparable to the findings of Copernicus Land Services and the Atlas of Urban Expansion with an underestimation of 1% and 8.75%, respectively. Improvements in overall accuracy (2.7%) and kappa coefficient (5%) were seen in classified maps from fused datasets. Fusion of Sentinel datasets provide a reliable means of impervious surface mapping at city scale as an indicator of environmental quality which is valuable for the sustainable management of the city.

Predicting the clinical management of skin lesions using deep learning

Automated machine learning approaches to skin lesion diagnosis from images are approaching dermatologist-level performance. However, current machine learning approaches that suggest management decisions rely on predicting the underlying skin condition to infer a management decision without considering the variability of management decisions that may exist within a single condition. We present the first work to explore image-based prediction of clinical management decisions directly without explicitly predicting the diagnosis. In particular, we use clinical and dermoscopic images of skin lesions along with patient metadata from the Interactive Atlas of Dermoscopy dataset (1011 cases; 20 disease labels; 3 management decisions) and demonstrate that predicting management labels directly is more accurate than predicting the diagnosis and then inferring the management decision (13.73 pm 3.93% and 6.59 pm 2.86% improvement in overall accuracy and AUROC respectively), statistically significant at p < 0.001. Directly predicting management decisions also considerably reduces the over-excision rate as compared to management decisions inferred from diagnosis predictions (24.56% fewer cases wrongly predicted to be excised). Furthermore, we show that training a model to also simultaneously predict the seven-point criteria and the diagnosis of skin lesions yields an even higher accuracy (improvements of 4.68 pm 1.89% and 2.24 pm 2.04% in overall accuracy and AUROC respectively) of management predictions. Finally, we demonstrate our model’s generalizability by evaluating on the publicly available MClass-D dataset and show that our model agrees with the clinical management recommendations of 157 dermatologists as much as they agree amongst each other.

Plant species classification in salt marshes using phenological parameters derived from Sentinel-2 pixel-differential time-series

Salt marshes are one of the most productive but vulnerable ecosystems on Earth; and due to the continued intensification of natural and anthropogenic pressures on them, accurate and timely information on the distribution of plant species in salt marshes is needed for effective coastal management. Time-series approaches have been widely applied to classify plant species; however, developing time-series with high spatial and temporal resolution over coastal zones remains challenging due to the influence caused by frequent cloud cover and periodic tidal fluctuations. In this study, aiming at the above challenges, we presented a saltmarsh vegetation classification method using phenological parameters derived from Sentinel-2 pixel-differential time-series (PDTS): first, a PDTS that each pixel has a different distribution of observations was constructed using a time-series cloud mask; second, a tidal filter determined by the threshold and frequency of the modified normalized difference water index (MNDWI) was used to exclude tide-related observations from the PDTS; third, phenological parameters that highlight the differences among salt marsh vegetation were extracted from a two-term Fourier fitting curve as classification features; and finally, the random forest algorithm was used for plant species classification with the assistance of sample data. Six common plant species (Spartina alterniflora, Phragmites australis, Suaeda salsa, Tamarix chinensis, Imperata cylindrica, and Scirpus mariqueter) from three representative coastal sites in China were analyzed. The major results were as follows: (1) The MNDWI demonstrated superior ability in identifying flooding pixels, with an overall accuracy of ~0.91. After tidal filtering, the R2 of the fitting curve for more than 70% of the vegetated salt marsh pixels was improved with an average increase of 0.113. (2) The six plant species exhibit unique phenological characteristics. In particular, P. australis has an advanced green-up season, 19–42 days earlier than that of the other plant species mentioned above, whereas S. alterniflora senesces one to two months later than the native plant species. (3) The average overall accuracy of the plant species classification based on the PDTS was 81.5%. Compared with a single-image-based classification, the PDTS-based classification demonstrated a ~ 5.1% improvement in overall accuracy, which is expected to serve the annual monitoring dynamics of the salt marsh.

Recent Developments in Instrumentation and its Application in Absolute Dating: Historical Perspective and Overview

The discovery of radioactivity in the early 20 th century led to the development of several radiometric dating methods (e.g., Rb-Sr, Sm-Nd, Re-Os, U-Pb, etc.). These radiometric dating methods are frequently used in earth science studies to constrain the deposition/formation timing of various natural archives (e.g., bulk rocks, minerals, carbonaceous materials , detrital clastic sedimentary materials, ore deposits , hydrocarbon deposits). The last few decades have witnessed significant improvements in overall accuracy and precision of these absolute radiometric dating methods due to continuous developments and refinements in sample processing and analytical techniques. In this contribution, we discuss some of the frequently used radiometric dating techniques for obtaining absolute ages in various natural archives and associated advancements in the instrumentation. The present attempt emphasizes on a multi-mineral and multi-isotopic approach with continuous developments in obtaining better precision and accuracy in the ages through improved analytical and measurement protocols that are the pre-requisite in absolute dating.

STransFuse: Fusing Swin Transformer and Convolutional Neural Network for Remote Sensing Image Semantic Segmentation

The applied research in remote sensing images has been pushed by convolutional neural network (CNN). Because of the fixed size of the perceptual field, CNN is unable to model global semantic relevance. Modeling global semantic information is possible with the self-attentive Transformer-based model. However, the method of patch computation used by Transformer for self-attentive computation ignores the spatial information inside each patch. To address these issues, we offer the STransFuse model as a new semantic segmentation method for remote sensing images. It is a model that combines the benefits of Transformer with CNN to improve the segmentation quality of various remote sensing images. We employ a staged model to extract coarse-grained and fine-grained feature representations at various semantic scales, unlike earlier techniques based on Transformer model fusion. In order to take full advantage of the features acquired at different stages, we designed an adaptive fusion module. This module adaptively fuses the semantic information between features at different scales employing a self-attentive mechanism. The overall accuracy (OA) of our proposed model on the Vaihingen dataset is 1.36% higher than the baseline, and 1.27% improvement in OA over baseline on the Potsdam dataset. When compared to other advanced models, the STransFuse model performs admirably.

SMOTEMultiBoost: Leveraging the SMOTE with MultiBoost to Confront the Class Imbalance in Supervised Learning

Class imbalance problem is being manifoldly confronted by researchers due to the increasing amount of complicated data. Common classification algorithms are impoverished to perform effectively on imbalanced datasets. Larger class cases typically outbalance smaller class cases in class imbalance learning. Common classification algorithms raise larger class performance owing to class imbalance in data and overall improvement in accuracy as their goal while lowering performance on smaller class. Furthermore, these algorithms deal false positive and false negative in an even way and regard equal cost of misclassifying cases. Meanwhile, different ensemble solutions have been proposed over the years for class imbalance learning but these approaches hamper the performance of larger class as emphasizing on the small class cases. The intuition of this overall degraded outcome would be the low diversity in ensemble solutions and overfitting or underfitting in data resampling techniques. To overcome these problems, we suggest a hybrid ensemble method by leveraging MultiBoost ensemble and Synthetic Minority Over-sampling TEchnique (SMOTE). Our suggested solution leverage the effectiveness of its elements. Therefore, it improves the outcome of the smaller class by reinforcing its space and limiting error in prediction. The proposed method shows improved performance as compare to numerous other algorithms and techniques in experiments.&#x0D;

Effective Training of Deep Convolutional Neural Networks for Hyperspectral Image Classification through Artificial Labeling

Hyperspectral imaging is a rich source of data, allowing for a multitude of effective applications. However, such imaging remains challenging because of large data dimension and, typically, a small pool of available training examples. While deep learning approaches have been shown to be successful in providing effective classification solutions, especially for high dimensional problems, unfortunately they work best with a lot of labelled examples available. The transfer learning approach can be used to alleviate the second requirement for a particular dataset: first the network is pre-trained on some dataset with large amount of training labels available, then the actual dataset is used to fine-tune the network. This strategy is not straightforward to apply with hyperspectral images, as it is often the case that only one particular image of some type or characteristic is available. In this paper, we propose and investigate a simple and effective strategy of transfer learning that uses unsupervised pre-training step without label information. This approach can be applied to many of the hyperspectral classification problems. The performed experiments show that it is very effective at improving the classification accuracy without being restricted to a particular image type or neural network architecture. The experiments were carried out on several deep neural network architectures and various sizes of labeled training sets. The greatest improvement in overall accuracy on the Indian Pines and Pavia University datasets is over 21 and 13 percentage points, respectively. An additional advantage of the proposed approach is the unsupervised nature of the pre-training step, which can be done immediately after image acquisition, without the need of the potentially costly expert’s time.

Transmission network reconstruction for foot-and-mouth disease outbreaks incorporating farm-level covariates.

Transmission network modelling to infer ‘who infected whom’ in infectious disease outbreaks is a highly active area of research. Outbreaks of foot-and-mouth disease have been a key focus of transmission network models that integrate genomic and epidemiological data. The aim of this study was to extend Lau’s systematic Bayesian inference framework to incorporate additional parameters representing predominant species and numbers of animals held on a farm. Lau’s Bayesian Markov chain Monte Carlo algorithm was reformulated, verified and pseudo-validated on 100 simulated outbreaks populated with demographic data Japan and Australia. The modified model was then implemented on genomic and epidemiological data from the 2010 outbreak of foot-and-mouth disease in Japan, and outputs compared to those from the SCOTTI model implemented in BEAST2. The modified model achieved improvements in overall accuracy when tested on the simulated outbreaks. When implemented on the actual outbreak data from Japan, infected farms that held predominantly pigs were estimated to have five times the transmissibility of infected cattle farms and be 49% less susceptible. The farm-level incubation period was 1 day shorter than the latent period, the timing of the seeding of the outbreak in Japan was inferred, as were key linkages between clusters and features of farms involved in widespread dissemination of this outbreak. To improve accessibility the modified model has been implemented as the R package ‘BORIS’ for use in future outbreaks.

Getting Rid of Dichotomous Sex Estimations: Why Logistic Regression Should be Preferred Over Discriminant Function Analysis.

Sex estimation is an important part of creating a biological profile for skeletal remains in forensics. The commonly used methods for developing sex estimation equations are discriminant function analysis (DFA) and logistic regression (LogR). LogR equations provide a probability of the predicted sex, while DFA relies on cutoff points to segregate males and females, resulting in a rigid dichotomization of the sexes. This is problematic because sexual dimorphism exists along a continuum and there can be considerable overlap in trait expression between the sexes. In this study, we used humeral measurements to compare the performance of DFA and LogR and found them to be very similar under multiple conditions. The overall cross‐validated (leave‐one‐out) accuracy of DFA (75.76–95.14%) was slightly higher than LogR (75.76–93.82%) for simple and multiple variable equations, and also performed better under varying sample sizes (94.03% vs. 93.78%). Three of five DFA equations outperformed LogR under the B index, while all five LogR equations outperformed the DFA equations under the Q index. Both methods saw an improvement in overall accuracy (DFA: 86.74–95.79%; LogR: 86.74–95.76%) when individuals with a classification probability lower than 0.80 were excluded. Additionally, we propose a method for calculating additional cutoff points (PMarks) based on posterior probability values. In conclusion, we recommend using LogR over DFA due to the increased flexibility, robusticity, and benefits for future users of the statistical models; however, if DFA is preferred, use of the proposed PMarks facilitates future analysis while avoiding unnecessary dichotomization.

Effects of Class Purity of Training Patch on Classification Performance of Crop Classification with Convolutional Neural Network

As the performance of supervised classification using convolutional neural networks (CNNs) are affected significantly by training patches, it is necessary to analyze the effects of the information content of training patches in patch-based classification. The objective of this study is to quantitatively investigate the effects of class purity of a training patch on performance of crop classification. Here, class purity that refers to a degree of compositional homogeneity of classes within a training patch is considered as a primary factor for the quantification of information conveyed by training patches. New quantitative indices for class homogeneity and variations of local class homogeneity over the study area are presented to characterize the spatial homogeneity of the study area. Crop classification using 2D-CNN was conducted in two regions (Anbandegi in Korea and Illinois in United States) with distinctive spatial distributions of crops and class homogeneity over the area to highlight the effect of class purity of a training patch. In the Anbandegi region with high class homogeneity, superior classification accuracy was obtained when using large size training patches with high class purity (7.1%p improvement in overall accuracy over classification with the smallest patch size and the lowest class purity). Training patches with high class purity could yield a better identification of homogenous crop parcels. In contrast, using small size training patches with low class purity yielded the highest classification accuracy in the Illinois region with low class homogeneity (19.8%p improvement in overall accuracy over classification with the largest patch size and the highest class purity). Training patches with low class purity could provide useful information for the identification of diverse crop parcels. The results indicate that training samples in patch-based classification should be selected based on the class purity that reflects the local class homogeneity of the study area.

DeepSuccinylSite: a deep learning based approach for protein succinylation site prediction

BackgroundProtein succinylation has recently emerged as an important and common post-translation modification (PTM) that occurs on lysine residues. Succinylation is notable both in its size (e.g., at 100 Da, it is one of the larger chemical PTMs) and in its ability to modify the net charge of the modified lysine residue from + 1 to − 1 at physiological pH. The gross local changes that occur in proteins upon succinylation have been shown to correspond with changes in gene activity and to be perturbed by defects in the citric acid cycle. These observations, together with the fact that succinate is generated as a metabolic intermediate during cellular respiration, have led to suggestions that protein succinylation may play a role in the interaction between cellular metabolism and important cellular functions. For instance, succinylation likely represents an important aspect of genomic regulation and repair and may have important consequences in the etiology of a number of disease states. In this study, we developed DeepSuccinylSite, a novel prediction tool that uses deep learning methodology along with embedding to identify succinylation sites in proteins based on their primary structure.ResultsUsing an independent test set of experimentally identified succinylation sites, our method achieved efficiency scores of 79%, 68.7% and 0.48 for sensitivity, specificity and MCC respectively, with an area under the receiver operator characteristic (ROC) curve of 0.8. In side-by-side comparisons with previously described succinylation predictors, DeepSuccinylSite represents a significant improvement in overall accuracy for prediction of succinylation sites.ConclusionTogether, these results suggest that our method represents a robust and complementary technique for advanced exploration of protein succinylation.

Design and Analysis of an Approximate Adder with Hybrid Error Reduction

This paper presents an energy-efficient approximate adder with a novel hybrid error reduction scheme to significantly improve the computation accuracy at the cost of extremely low additional power and area overheads. The proposed hybrid error reduction scheme utilizes only two input bits and adjusts the approximate outputs to reduce the error distance, which leads to an overall improvement in accuracy. The proposed design, when implemented in 65-nm CMOS technology, has 3, 2, and 2 times greater energy, power, and area efficiencies, respectively, than conventional accurate adders. In terms of the accuracy, the proposed hybrid error reduction scheme allows that the error rate of the proposed adder decreases to 50% whereas those of the lower-part OR adder and optimized lower-part OR constant adder reach 68% and 85%, respectively. Furthermore, the proposed adder has up to 2.24, 2.24, and 1.16 times better performance with respect to the mean error distance, normalized mean error distance (NMED), and mean relative error distance, respectively, than the other approximate adder considered in this paper. Importantly, because of an excellent design tradeoff among delay, power, energy, and accuracy, the proposed adder is found to be the most competitive approximate adder when jointly analyzed in terms of the hardware cost and computation accuracy. Specifically, our proposed adder achieves 51%, 49%, and 47% reductions of the power-, energy-, and error-delay-product-NMED products, respectively, compared to the other considered approximate adders.

POL-SAR Image Classification Based on Modified Stacked Autoencoder Network and Data Distribution

This article proposes a novel autoencoder (AE) network based on the distribution of polarimetric synthetic aperture radar (POL-SAR) data matrix, called a mixture autoencoder (MAE). Through a detailed analysis of the data distribution POL-SAR data matrix, a normalization method is also presented in succession. The proposed MAE defines the data error term in the loss function according to the data distribution. It can be regarded as a process of unsupervised feature extraction designed specifically for POL-SAR data matrix. Then, a softmax classifier is trained with the help of data features and the corresponding label information. Next, a stacked MAE (SMAE) network is reasonably constructed by considering the data distribution among different layers. Finally, this article also presents a classification network through discarding the decoder process of the proposed SMAE and connecting with a softmax classifier. The SMAE is trained layer by layer using the unlabeled data. The softmax classifier is also trained with a small number of labeled pixels. With parameters obtained from the above-mentioned procedures as the initial parameters, the whole classification network is trained by the labeled pixels to get a well-trained model, which is used for predicting the corresponding label of the pixel in the data set. Three real POL-SAR data sets, including the AIR-SAR L-band data of Flevoland, The Netherlands, are used in the experiments. Compared with one classical algorithm and two related models with the similar structure, both the proposed methods show improvements in overall accuracy and efficiency as well as possess better adaptability of the parameter and preferable consistency with the classification performance.

Wetland Classification Based on a New Efficient Generative Adversarial Network and Jilin-1 Satellite Image

Recent studies have shown that deep learning methods provide useful tools for wetland classification. However, it is difficult to perform species-level classification with limited labeled samples. In this paper, we propose a semi-supervised method for wetland species classification by using a new efficient generative adversarial network (GAN) and Jilin-1 satellite image. The main contributions of this paper are twofold. First, the proposed method, namely ShuffleGAN, requires only a small number of labeled samples. ShuffleGAN is composed of two neural networks (i.e., generator and discriminator), which perform an adversarial game in the training phase and ShuffleNet units are added in both generator and discriminator to obtain speed-accuracy tradeoff. Second, ShuffleGAN can perform species-level wetland classification. In addition to distinguishing the wetland areas from non-wetlands, different tree species located in the wetland are also identified, thus providing a more detailed distribution of the wetland land-covers. Experiments are conducted on the Haizhu Lake wetland data acquired by the Jilin-1 satellite. Compared with existing GAN, the improvement in overall accuracy (OA) of the proposed ShuffleGAN is more than 2%. This work can not only deepen the application of deep learning in wetland classification but also promote the study of fine classification of wetland land-covers.

CNN-BASED FEATURE-LEVEL FUSION OF VERY HIGH RESOLUTION AERIAL IMAGERY AND LIDAR DATA

Abstract. Land-cover classification of Remote Sensing (RS) data in urban area has always been a challenging task due to the complicated relations between different objects. Recently, fusion of aerial imagery and light detection and ranging (LiDAR) data has obtained a great attention in RS communities. Meanwhile, convolutional neural network (CNN) has proven its power in extracting high-level (deep) descriptors to improve RS data classification. In this paper, a CNN-based feature-level framework is proposed to integrate LiDAR data and aerial imagery for object classification in urban area. In our method, after generating low-level descriptors and fusing them in a feature-level fusion by layer-stacking, the proposed framework employs a novel CNN to extract the spectral-spatial features for classification process, which is performed using a fully connected multilayer perceptron network (MLP). The experimental results revealed that the proposed deep fusion model provides about 10% improvement in overall accuracy (OA) in comparison with other conventional feature-level fusion techniques.

Long-short-term-memory-based crop classification using high-resolution optical images and multi-temporal SAR data

Farmland parcel-based crop classification using satellite data plays an important role in precision agriculture. In this study, a deep-learning-based time-series analysis method employing optical images and synthetic aperture radar (SAR) data is presented for crop classification for cloudy and rainy regions. Central to this method is the spatial-temporal incorporation of high-resolution optical images and multi-temporal SAR data and deep-learning-based time-series analysis. First, a precise farmland parcel map is delineated from high-resolution optical images. Second, pre-processed SAR intensity images are overlaid onto the parcel map to construct time series of crop growth for each parcel. Third, a deep-learning-based (using the long short-term memory, LSTM, network) classifier is employed to learn time-series features of crops and to classify parcels to produce a final classification map. The method was applied to two datasets of high-resolution ZY-3 images and multi-temporal Sentinel-1A SAR data to classify crop types in Hunan and Guizhou of China. The classification results, with an 5.0% improvement in overall accuracy compared to those of traditional methods, illustrate the effectiveness of the proposed framework for parcel-based crop classification for southern China. A further analysis of the relationship between crop calendars and change patterns of time-series intensity indicates that the LSTM model could learn and extract useful features for time-series crop classification.

Auto Encoder Feature Learning with Utilization of Local Spatial Information and Data Distribution for Classification of PolSAR Image

The distribution of data plays a key role in the designing of a machine learning model. Therefore, this paper proposes a novel auto encoder network based on the distribution of polarimetric synthetic aperture radar (PolSAR) data matrix. Designed specifically for PolSAR data matrix, the proposed mixture auto encoder (MAE) feature learning method defines data error term in the loss function according to the data distribution. Instead of the pixel itself, all pixels in the neighborhood are used as input to train the proposed MAE. Then, a corresponding classification network is also given by discarding the decoder process of the proposed MAE and connecting with a Softmax classifier. The MAE is trained using the unlabeled data, while the training process of the classification network is completed with the help of a small number of labeled pixels. In view of the phenomenon of misclassification in the predicted result image, two post-processing steps acting on local spatial are also given, which accomplished by the proposed two filters. Extensive experiments by four methods were made over three real PolSAR images including the proposed classification network. The experimental results show that introducing data distribution into the auto encoder network leads to an average 4% improvement in overall accuracy for three PolSAR images. Moreover, the post-processing steps with the proposed filters bring a new level of discrimination on the classification performance of PolSAR images.

Assessment of Optimal Transport for Operational Land-Cover Mapping Using High-Resolution Satellite Images Time Series without Reference Data of the Mapping Period

Land-cover map production using remote-sensing imagery is governed by data availability. In our case, data sources are two-fold: on one hand, optical data provided regularly by satellites such as Sentinel-2, and on the other hand, reference data which allow calibrating mapping methods or validating the results. The lengthy delays due to reference data collection and cleansing are one of the main issues for applications. In this work, the use of Optimal Transport (OT) is proposed. OT is a Domain Adaptation method that uses past data, both images and reference data, to produce the land-cover map of the current period without updated reference data. Seven years of Formosat-2 image time series and the corresponding reference data are used to evaluate two OT algorithms: conventional EMD transport and regularized transport based on the Sinkhorn distance. The contribution of OT to a classification fusion strategy is also evaluated. The results show that with a 17-class nomenclature the problem is too complex for the Sinkhorn algorithm, which provides maps with an Overall Accuracy (OA) of 30%. In contrast, with the EMD algorithm, an OA close to 70% is obtained. One limitation of OT is the number of classes that can be considered at the same time. Simplification schemes are proposed to reduce the number of classes to be transported. Cases of improvement are shown when the problem is simplified, with an improvement in OA varying from 5% and 20%, producing maps with an OA near 79%. As several years are available, the OT approaches are compared to standard fusion schemes, like majority voting. The gain in voting strategies with OT use is lower than the gain obtained with standard majority voting (around 5%).

Effect of introducing weather parameters on the accuracy of milk production forecast models

The objective of this study was to analyze the effect of adding meteorological data to the training process of two milk production forecast models. The two models chosen were the nonlinear auto-regressive model with exogenous input (NARX) and the multiple linear regression (MLR) model. The accuracy of these models were assessed using seven different combinations of precipitation, sunshine hours and soil temperature as additional model training inputs. Lactation data (daily milk yield and days in milk) from 39 pasture-based Holstein-Friesian Irish dairy cows were selected to compare to the model outputs from a central database. The models were trained using historical milk production data from three lactation cycles and were employed to predict the total daily milk yield of a fourth lactation cycle for each individual cow over short (10-day), medium (30-day) and long-term (305-day) forecast horizons. The NARX model was found to provide a greater prediction accuracy when compared to the MLR model when predicting annual individual cow milk yield (kg), with R2 values greater than 0.7 for 95.5% and 14.7% of total predictions, respectively. The results showed that the introduction of sunshine hours, precipitation and soil temperature data improved the prediction accuracy of individual cow milk prediction for the NARX model in the short, medium and long-term forecast horizons. Sunshine hours was shown to have the largest impact on milk production with an improvement of forecast accuracy observed in 60% and 70% of all predictions (for all 39 test cows from both groups). However, the overall improvement in accuracy was small with a maximum forecast error reduction of 4.3%. Thus, the utilization of meteorological parameters in milk production forecasting did not have a substantial impact on forecast accuracy.

Contextual Action Cues from Camera Sensor for Multi-Stream Action Recognition.

In action recognition research, two primary types of information are appearance and motion information that is learned from RGB images through visual sensors. However, depending on the action characteristics, contextual information, such as the existence of specific objects or globally-shared information in the image, becomes vital information to define the action. For example, the existence of the ball is vital information distinguishing “kicking” from “running”. Furthermore, some actions share typical global abstract poses, which can be used as a key to classify actions. Based on these observations, we propose the multi-stream network model, which incorporates spatial, temporal, and contextual cues in the image for action recognition. We experimented on the proposed method using C3D or inflated 3D ConvNet (I3D) as a backbone network, regarding two different action recognition datasets. As a result, we observed overall improvement in accuracy, demonstrating the effectiveness of our proposed method.