Multilayer Perceptron Architecture Research Articles

Artificial neural networks modeling of combustion parameters for a diesel engine fueled with biodiesel fuel

In the present study, numerous artificial neural networks were employed to predict the combustion characteristics of a four-stroke, single-cylinder, naturally aspirated diesel engine, including multilayer perceptron (MLP), adaptive neuro-fuzzy interference system (ANFIS) and radial basis function network (RBFN). The actual data derived from measurements and calculations were applied in model training, cross-validation, and testing. Biodiesel fuel ratio, engine load, air consumption, and fuel flow rate data were considered as model-input parameters, which are related to main engine operating variables and also affect the combustion characteristics. These kinds of model-input data were especially preferred due to being commonly direct measurable with the main engine sensors or found in engine maps/look-up tables managed by the electronic control unit (ECU). The main parameters obtained from the direct analysis of the measured in-cylinder pressure data and the heat release analysis results were also determined as model-output parameters. Equally, to ensure a more accurate, straightforward and practical approach to prediction, in every category of neural networks, three training algorithms were adopted, including Levenberg-Marquardt (LM), back propagation (BP) and conjugate gradient (CG). Moreover, a sensitivity analysis was accomplished by assessing the strengths of the links between input and output parameters for every topology. As such, the researcher revealed that all proposed neural networks could predict maximum heat release rate (HRRmax), ignition delay (ID), maximum cylinder pressure (Pmax), maximum cylinder pressure location (θPmax), maximum in-cylinder pressure rise rate (dPmax), indicated mean effective pressure (IMEP), crank angle of center heat release rate (CA50), and combustion duration (CD), with a high accuracy rate. Results of model-output parameters are also important parameters in the combustion diagnostics which are influential on engine thermal efficiency and pollutant formation process. In addition, it is necessary to note that MLP architectures that incorporated the LM algorithm presented superior results. With regards to the optimal ANN model, the linear coefficient values: 0.999848, 0.999847, 0.999955, 0.999780, 0.999378, 0.999929, 0.999766, and 0.999216, were found for ID, HRRmax, Pmax, θPmax, dPmax, IMEP, CA50 and CD, respectively.

МОДЕЛЬ ПРОГНОЗИРОВАНИЯ ТРАНСПОРТНОГО ПОТОКА НА ОСНОВЕ НЕЙРОННЫХ СЕТЕЙ ДЛЯ ПРЕДСКАЗАНИЯ ТРАФИКА НА ДОРОГАХ

In connection with the industrialization of modern society, the growth of the transport systemsof our country, an increase in certain necessary for the development of the needs of the citizensof our country, the number of vehicles of various types continues to increase every year veryfast, causing huge traffic jams on transport roads, especially in large cities and megacities. Thus,forecasting traffic flows is an important and necessary component of optimal traffic control inmodern conditions of transport network development. As a solution to this problem, this articleaims to analyze and describe the application of artificial intelligence methods, in particular neuralnetworks, which represent a modern approach to modeling in complex and nonlinear situationsthat arise when predicting a traffic flow model. The shown accuracy method is based on the developmentof a neural network to predict the daily traffic flow. The expected traffic flow is then comparedwith the actual dataset recorded on the road section and provided by the infrastructuremanager. In fact, neural networks are able to learn from past situations and predict future situationson the transport network. In this study, various neural network structures were examined, and the simulation results showed that the best predictions were obtained using the multilayerperceptron architecture, which has a good generalization system with a root mean square error of0.00927 with the current set of vehicles. The first part of the article is devoted to defining variousconcepts related to the current research area, including a review of the literature on traffic predictionand neural networks. The second part is devoted to describing the problem of traffic congestionusing forecasting problems and presenting the proposed solution method with an emphasis onartificial neural networks as a means of forecasting demand and its various structures. Then, numericalexperiments are illustrated by analyzing the forecast results after the formation and testingof various neural network architectures.

Rice-Fusion: A Multimodality Data Fusion Framework for Rice Disease Diagnosis

Rice leaf infections are a common hazard to rice production, affecting many farmers all over the world. Early detection and treatment of rice leaf infection are critical for promoting healthy rice plant growth and ensuring adequate supply for the fast-growing population. Computer-assisted rice leaf disease diagnoses are hampered due to strong image backgrounds. Popular Convolutional Neural Network (CNN) architecture extracts the features from images and diagnoses the disease to address the issues above. However, this method is best suitable for segmented images and gives low accuracy with real-time images. In this case, the Internet of Things is a paradigm shift that collects agro-meteorological information that effectively helps diagnose rice diseases. Motivated by the usefulness of CNN models and agricultural IoT, a novel multimodal data fusion framework named Rice-Fusion is proposed to diagnose rice disease. Rice disease diagnosis based on a single modality may not be accurate, and hence the fusion of heterogeneous modalities is essential for robust and reliable disease diagnosis. This gives a new dimension to the domain of rice disease diagnosis. The dataset was collected manually with 3200 rice health category samples using two modalities, namely agro-meteorological sensors and a camera. The Rice-Fusion framework initially extracts the numerical features from agro-meteorological data collected from sensors. Next, it extracts the visual features from the captured rice images. These extracted features are further fused using a concatenation layer followed by a dense layer, which provides single output for diagnosing the rice disease. The testing accuracy of Rice-Fusion is 95.31% as opposed to other unimodal framework accuracies of 82.03% and 91.25% based on CNN and Multi-Layer Perceptron (MLP) architectures, respectively. Experimental results analysis demonstrates that the proposed Rice-Fusion multimodal data fusion framework outperforms the outcome of unimodal frameworks.

Design of an Intelligent Hydroponics System to Identify Macronutrient Deficiencies in Chili

Nutrient contents are important for plants. Lack of macronutrients causes plant damage. Several macronutrient deficiencies exhibit similar visual characteristics that are difficult for ordinary farmers to identify. Collaboration between Computer Vision technology and IoT has become a non-destructive method for nutrient monitoring and control, included in the hydroponic system. Computer vision plays a role in processing plant image data based on specific characteristics. However, the analysis of one characteristic cannot represent plant health. In addition, knowing the percentage of macronutrient deficiencies is also needed to support precision agriculture systems. Therefore, we propose a Multi Layer Perceptron architecture that can perform multi-tasks, namely, identification and estimation. In addition, the optimal architecture will also be sought based on the characteristics of the combination of three features in the form of texture, color, and leaf shape. Based on analysis and design, our proposed model has a high potential for identifying and estimating macronutrient deficiency at the same time as well and can be applied to support precision agriculture in Indonesia.

Multilayer Perceptron Generative Model via Adversarial Learning for Robust Visual Tracking

Visual tracking is an open and exciting field of research. The researchers introduced great efforts to be close to the ideal state of stable tracking of objects regardless of different appearances or circumstances. Owing to the attractive advantages of generative adversarial networks (GANs), they have been a promising area of research in many fields. However, GAN network architecture has not been thoroughly investigated in the visual tracking research community. Inspired by visual tracking via adversarial learning (VITAL), we present a novel network to generate randomly initialized masks for building augmented feature maps using multilayer perceptron (MLP) generative models. To obtain more robust tracking these augmented masks can extract robust features that do not change over a long temporal span. Some models such as deep convolutional generative adversarial networks (DCGANs) have been proposed to obtain powerful generator architectures by eliminating or minimizing the use of fully connected layers. This study demonstrates that the use of MLP architecture for the generator is more robust and efficient than the convolution-only architecture. Also, to realize better performance, we used one-sided label smoothing to regularize the discriminator in the training stage and the label smoothing regularization (LSR) method to reduce the overfitting of the classifier in the online tracking stage. The experiments show that the proposed model is more robust than the DCGAN model and offers satisfactory performance compared with the state-of-the-art deep visual trackers on OTB-100, VOT2019 and LaSOT datasets.

A comparison of conventional and deep learning methods of image classification

The aim of the research is to compare traditional and deep learning methods in image classification tasks. The conducted research experiment covers the analysis of five different models of neural networks: two models of multi–layer perceptron architecture: MLP with two hidden layers, MLP with three hidden layers; and three models of convolutional architecture: the three VGG blocks model, AlexNet and GoogLeNet. The models were tested on two different datasets: CIFAR–10 and MNIST and have been applied to the task of image classification. They were tested for classification performance, training speed, and the effect of the complexity of the dataset on the training outcome.

A Multi-Layer Perceptron Model for Classification of E-mail Fraud

This study was developed an e-mail classification model to preempt fraudulent activities. The e-mail has such a predominant nature that makes it suitable for adoption by cyber-fraudsters. This research used a combination of two databases: CLAIR fraudulent and Spambase datasets for creating the training and testing dataset. The CLAIR dataset consists of raw e-mails from users’ inbox which were pre-processed into structured form using Natural Language Processing (NLP) techniques. This dataset was then consolidated with the Spambase dataset as a single dataset. The study deployed the Multi-Layer Perceptron (MLP) architecture which used a back-propagation algorithm for training the fraud detection model. The model was simulated using 70% and 80% for training while 30% and 20% of datasets were used for testing respectively. The results of the performance of the models were compared using a number of evaluation metrics. The study concluded that using the MLP, an effective model for fraud detection among e-mail dataset was proposed.

Deep Learning Methods for Vessel Trajectory Prediction Based on Recurrent Neural Networks

Data-driven methods open up unprecedented possibilities for maritime surveillance using Automatic Identification System (AIS) data. In this work, we explore deep learning strategies using historical AIS observations to address the problem of predicting future vessel trajectories with a prediction horizon of several hours. We propose novel sequence-to-sequence vessel trajectory prediction models based on encoder-decoder recurrent neural networks (RNNs) that are trained on historical trajectory data to predict future trajectory samples given previous observations. The proposed architecture combines Long Short-Term Memory (LSTM) RNNs for sequence modeling to encode the observed data and generate future predictions with different intermediate aggregation layers to capture space-time dependencies in sequential data. Experimental results on vessel trajectories from an AIS dataset made freely available by the Danish Maritime Authority show the effectiveness of deep-learning methods for trajectory prediction based on sequence-to-sequence neural networks, which achieve better performance than baseline approaches based on linear regression or on the Multi-Layer Perceptron (MLP) architecture. The comparative evaluation of results shows: i) the superiority of attention pooling over static pooling for the specific application, and ii) the remarkable performance improvement that can be obtained with labeled trajectories, i.e., when predictions are conditioned on a low-level context representation encoded from the sequence of past observations, as well as on additional inputs (e.g., port of departure or arrival) about the vessel's high-level intention, which may be available from AIS.

Individual outcome prediction models for patients with COVID-19 based on their first day of admission to the intensive care unit

BackgroundCurrently, good prognosis and management of critically ill patients with COVID-19 are crucial for developing disease management guidelines and providing a viable healthcare system. We aimed to propose individual outcome prediction models based on binary logistic regression (BLR) and artificial neural network (ANN) analyses of data collected in the first 24 h of intensive care unit (ICU) admission for patients with COVID-19 infection. We also analysed different variables for ICU patients who survived and those who died. MethodsData from 326 critically ill patients with COVID-19 were collected. Data were captured on laboratory variables, demographics, comorbidities, symptoms and hospital stay related information. These data were compared with patient outcomes (survivor and non-survivor patients). BLR was assessed using the Wald Forward Stepwise method, and the ANN model was constructed using multilayer perceptron architecture. ResultsThe area under the receiver operating characteristic curve of the ANN model was significantly larger than the BLR model (0.917 vs 0.810; p < 0.001) for predicting individual outcomes. In addition, ANN model presented similar negative predictive value than the BLR model (95.9% vs 94.8%).Variables such as age, pH, potassium ion, partial pressure of oxygen, and chloride were present in both models and they were significant predictors of death in COVID-19 patients. ConclusionsOur study could provide helpful information for other hospitals to develop their own individual outcome prediction models based, mainly, on laboratory variables. Furthermore, it offers valuable information on which variables could predict a fatal outcome for ICU patients with COVID-19.

Dataset Generation for Development of Multi-Node Cyber Threat Detection Systems

This paper presents a new approach to generate datasets for cyber threat research in a multi-node system. For this purpose, the proof-of-concept of such a system is implemented. The system will be used to collect unique datasets with examples of information hiding techniques. These techniques are not present in publicly available cyber threat detection datasets, while the cyber threats that use them represent an emerging cyber defense challenge worldwide. The network data were collected thanks to the development of a dedicated application that automatically generates random network configurations and runs scenarios of information hiding techniques. The generated datasets were used in the data-driven research workflow for cyber threat detection, including the generation of data representations (network flows), feature selection based on correlations, data augmentation of training datasets, and preparation of machine learning classifiers based on Random Forest and Multilayer Perceptron architectures. The presented results show the usefulness and correctness of the design process to detect information hiding techniques. The challenges and research directions to detect cyber deception methods are discussed in general in the paper.

Similarity-Based Virtual Screen Using Enhanced Siamese Multi-Layer Perceptron.

Traditional drug development is a slow and costly process that leads to the production of new drugs. Virtual screening (VS) is a computational procedure that measures the similarity of molecules as one of its primary tasks. Many techniques for capturing the biological similarity between a test compound and a known target ligand have been established in ligand-based virtual screens (LBVSs). However, despite the good performances of the above methods compared to their predecessors, especially when dealing with molecules that have structurally homogenous active elements, they are not satisfied when dealing with molecules that are structurally heterogeneous. The main aim of this study is to improve the performance of similarity searching, especially with molecules that are structurally heterogeneous. The Siamese network will be used due to its capability to deal with complicated data samples in many fields. The Siamese multi-layer perceptron architecture will be enhanced by using two similarity distance layers with one fused layer, then multiple layers will be added after the fusion layer, and then the nodes of the model that contribute less or nothing during inference according to their signal-to-noise ratio values will be pruned. Several benchmark datasets will be used, which are: the MDL Drug Data Report (MDDR-DS1, MDDR-DS2, and MDDR-DS3), the Maximum Unbiased Validation (MUV), and the Directory of Useful Decoys (DUD). The results show the outperformance of the proposed method on standard Tanimoto coefficient (TAN) and other methods. Additionally, it is possible to reduce the number of nodes in the Siamese multilayer perceptron model while still keeping the effectiveness of recall on the same level.

SS-MLP: A Novel Spectral-Spatial MLP Architecture for Hyperspectral Image Classification

Convolutional neural networks (CNNs) are the go-to model for hyperspectral image (HSI) classification because of the excellent locally contextual modeling ability that is beneficial to spatial and spectral feature extraction. However, CNNs with a limited receptive field pose challenges for modeling long-range dependencies. To solve this issue, we introduce a novel classification framework which regards the input HSI as a sequence data and is constructed exclusively with multilayer perceptrons (MLPs). Specifically, we propose a spectral-spatial MLP (SS-MLP) architecture, which uses matrix transposition and MLPs to achieve both spectral and spatial perception in global receptive field, capturing long-range dependencies and extracting more discriminative spectral-spatial features. Four benchmark HSI datasets are used to evaluate the classification performance of the proposed SS-MLP. Experimental results show that our pure MLP-based architecture outperforms other state-of-the-art convolution-based models in terms of both classification performance and computational time. When comparing with the SSSERN model, the average accuracy improvement of our approach is as high as 3.03%. We believe that our impressive experimental results will foster additional research on simple yet effective MLP-based architecture for HSI classification.

Modifications of the Multi-Layer Perceptron for Hyperspectral Image Classification

Recently, many convolutional neural network (CNN)-based methods have been proposed to tackle the classification task of hyperspectral images (HSI). In fact, CNN has become the de-facto standard for HSI classification. It seems that the traditional neural networks such as multi-layer perceptron (MLP) are not competitive for HSI classification. However, in this study, we try to prove that the MLP can achieve good classification performance of HSI if it is properly designed and improved. The proposed Modified-MLP for HSI classification contains two special parts: spectral–spatial feature mapping and spectral–spatial information mixing. Specifically, for spectral–spatial feature mapping, each input sample of HSI is divided into a sequence of 3D patches with fixed length and then a linear layer is used to map the 3D patches to spectral–spatial features. For spectral–spatial information mixing, all the spectral–spatial features within a single sample are feed into the solely MLP architecture to model the spectral–spatial information across patches for following HSI classification. Furthermore, to obtain the abundant spectral–spatial information with different scales, Multiscale-MLP is proposed to aggregate neighboring patches with multiscale shapes for acquiring abundant spectral–spatial information. In addition, the Soft-MLP is proposed to further enhance the classification performance by applying soft split operation, which flexibly capture the global relations of patches at different positions in the input HSI sample. Finally, label smoothing is introduced to mitigate the overfitting problem in the Soft-MLP (Soft-MLP-L), which greatly improves the classification performance of MLP-based method. The proposed Modified-MLP, Multiscale-MLP, Soft-MLP, and Soft-MLP-L are tested on the three widely used hyperspectral datasets. The proposed Soft-MLP-L leads to the highest OA, which outperforms CNN by 5.76%, 2.55%, and 2.5% on the Salinas, Pavia, and Indian Pines datasets, respectively. The obtained results reveal that the proposed models provide competitive results compared to the state-of-the-art methods, which shows that the MLP-based methods are still competitive for HSI classification.

Arabic Personal Name Matching: Names Written using Latin Alphabet

: In many Arab countries’ public administrations, Arabic personal names are written with Latin alphabet, generally, in various ways by different writers. This has led to many problems when it comes to connecting these administrations. The aim of this study was to propose two new approaches for the pairwise matching of Arabic personal names. The first approach is based on string alignment and phonetic transcription. Appropriate scoring functions were defined to catch similarity between Arabic personal names. In the second approach, we use machine learning techniques to derive a suitable model for this problem. Precisely, we suggest using a Multi-Layer Perceptron (MLP) architecture and experiment with different configurations. Performances of the new models compare well with the best-performing similarity measures (Jaro, Jaro-Winkler, Double Metaphone and Edit Distance) in terms of precision, recall and F1. Even though the work was carried out for the (Algeria/French Alphabet) case, it can be adapted to any other (country/script) case, like (Egypt/English).

On various multi-layer perceptron and radial basis function based artificial neural networks in the process of a hot flow curve description

In recent years, the study of the hot deformation behavior of various materials is significantly marked by an increasing utilization of artificial neural networks, which are frequently employed for a hot flow curve description. This specific kind of description is commonly solved via a Feed-Forward Multi-Layer Perceptron architecture and rarely via a Radial Basis architecture. Both network architectures are compared to assess their suitability in the process of a hot flow curve description under a wide range of thermomechanical conditions. The performed survey is also aimed on the eventual utilization of corresponding modifications of both studied networks, namely on a Cascade-Forward Multi-Layer Perceptron and Generalized Regression network. The main results have shown that the Feed-Forward Multi-Layer Perceptron architecture represents a good choice if very high accuracy is a crucial goal. However, in the case of this architecture, finding the proper parameters can be time-consuming and the hardware burdensome. On the contrary, for the flow curve description the almost unused Radial Basis network offers a very easy training procedure and significantly shorter computing time under acceptable accuracy. The results of the submitted research should then serve as a background for the selection and following application of a suitable network architecture in the process of solving future flow curve description tasks.

Reactive Obstacle–Avoidance Systems for Wheeled Mobile Robots Based on Artificial Intelligence

Obstacle–Avoidance robots have become an essential field of study in recent years. This paper analyzes two cases that extend reactive systems focused on obstacle detection and its avoidance. The scenarios explored get data from their environments through sensors and generate information for the models based on artificial intelligence to obtain a reactive decision. The main contribution is focused on the discussion of aspects that allow for comparing both approaches, such as the heuristic approach implemented, requirements, restrictions, response time, and performance. The first case presents a mobile robot that applies a fuzzy inference system (FIS) to achieve soft turning basing its decision on depth image information. The second case introduces a mobile robot based on a multilayer perceptron (MLP) architecture, which is a class of feedforward artificial neural network (ANN), and ultrasonic sensors to decide how to move in an uncontrolled environment. The analysis of both options offers perspectives to choose between reactive Obstacle–Avoidance systems based on ultrasonic or Kinect sensors, models that infer optimal decisions applying fuzzy logic or artificial neural networks, with key elements and methods to design mobile robots with wheels. Therefore, we show how AI or Fuzzy Logic techniques allow us to design mobile robots that learn from their “experience” by making them safe and adjustable for new tasks, unlike traditional robots that use large programs to perform a specific task.

Artificial intelligence in percutaneous coronary intervention: improved risk prediction of PCI-related complications using an artificial neural network

ObjectivesComplications after percutaneous coronary intervention (PCI) are common and costly. Risk models for predicting the likelihood of acute kidney injury (AKI), bleeding, stroke and death are limited by accuracy and...

Comprehensive proteomic analysis reveals distinct functional modules associated with skull base and supratentorial meningiomas and perturbations in collagen pathway components

Meningiomas are brain tumors that originate from the meninges and has been primarily classified into three grades by the current WHO guidelines. Although widely prevalent and can be managed by surgery there are instances when the tumors are located in difficult regions. This results in considerable challenges for complete surgical resection and further clinical management. While the genetic signature of the skull base tumors is now known to be different from the non-skull base tumors, there is a lack of information at the functional aspects of these tumors at the proteomic level. Thus, the current study thereby aims to obtain mechanistic insights between the two radiologically distinct groups of meningiomas, namely the skull base & supratentorial (non-skull base-NSB) regions. We have employed a comprehensive mass spectrometry-based label-free quantitative proteomic analysis in Skull base and supratentorial meningiomas. Further, we have used an Artificial Neural Networking employing a sparse Multilayer perceptron (MLP) architecture to predict protein concordance. A patient-derived spectral library has been employed for a novel peptide-level validation of proteins that are specific to the radiological regions using the SRM assay based targeted proteomics approach. The comprehensive proteomics enabled the identification of nearly 4000 proteins with high confidence (1%FDR ≥ 2 unique peptides) among which 170 proteins were differentially abundant in Skull base vs Supratentorial tumors (p-value ≤0.05). In silico analysis enabled mapping of the major alterations and hinted towards an overall perturbation of extracellular matrix and collagen biosynthesis components in the non-skull base meningiomas and a prominent perturbation of molecular trafficking in the skull base meningiomas. Therefore, this study has yielded novel insights into the functional association of the proteins that are differentially abundant in the two radiological subgroups. SignificanceIn the current study, we have performed label-free proteomic analysis on fresh frozen tissue of 14 Supratentorial (NSB) and 7 Skull base meningiomas to assess perturbations in the global proteome, we have further employed an in-depth in silico analysis to map the pathways that have enabled functional mapping of the differentially abundant proteins in the Skull base and Supratentorial tumors. The findings from the above were also subjected to a machine learning-based neural networking to find out the proteins that have the most concordance of occurrence to determine the most influential proteins of the network. We further validated the differential abundance of identified protein markers in a larger patient cohort of Skull base and Supratentorial employing targeted proteomics approach to validate key protein candidates emerging from ours and other recent studies.The previous studies that have explored the skull base and convexity meningiomas have been able to reveal alterations in the genetic mutations in these tumor types. However, there are not many studies that have explored the functional aspects of these tumors, especially at the proteome level. We have attempted for the first time to map the functional modules associated with altered proteins in these tumors and have been able to identify that there is a possibility that the Skull base meningiomas to be considerably different from the Non-skull base (NSB) tumors in terms of the perturbed pathways. Our study employed global as well as targeted proteomics to examine the proteomic alterations in these two tumor groups. The study indicates that proteins that were more abundant in Skull base tumors were part of molecular transport components, non-skull base proteins majorly mapped to the components of extracellular matrix remodeling pathways. In conclusion, this study substantiates the distinction in the proteomic signatures in the skull base and supratentorial meningiomas paving way for further investigation of the identified markers for determining if some of these proteins can be used for therapeutic interventions for cases that pose considerable challenges for complete resection.

High-Density Reconfigurable Synaptic Transistors Targeting a Minimalist Neural Network.

Enormous synaptic devices are required to build a parallel, precise, and efficient neural computing system. To further improve the energy efficiency of neuromorphic computing, a single high-density synaptic (HDS) device with multiple nonvolatile synaptic states is suggested to reduce the number of synaptic devices in the neural network, although such a powerful synaptic device is rarely demonstrated. Here, a photoisomerism material, namely, diarylethene, whose energy level varies with the wavelength of illumination is first introduced to construct a powerful HDS device. The multiple synaptic states of the HDS device are intrinsically converted under UV-vis regulation and remain nonvolatile after the removal of illumination. More importantly, the conversion is reconfigurable and reversible under different light conditions, and the synaptic characteristics are comprehensively mimicked in each state. Finally, compared with a two-layer multilayer perceptron (MLP) architecture based on static synaptic devices, the HDS device-based architecture reduces the device number by 16 times to achieve a minimalist neural computing structure. The invention of the HDS device opens up a revolutionary paradigm for the establishment of a brain-like network.

Deep learning strategies for automatic fault diagnosis in photovoltaic systems by thermographic images

Losses of electricity production in photovoltaic systems are mainly caused by the presence of faults that affect the efficiency of the systems. The identification of any overheating in a photovoltaic module, through the thermographic non-destructive test, may be essential to maintain the correct functioning of the photovoltaic system quickly and cost-effectively, without interrupting its normal operation. This work proposes a system for the automatic classification of thermographic images using a convolutional neural network, developed via open-source libraries. To reduce image noise, various pre-processing strategies were evaluated, including normalization and homogenization of pixels, greyscaling, thresholding, discrete wavelet transform, and Sobel Feldman and box blur filtering. These techniques allow the classification of thermographic images of differen quality and acquired using different equipments, without specific protocols. Several tests with different parameters and overfitting reduction techniques were carried out to assess the performance of the neural networks: images acquired by unmanned aerial vehicles and ground-based operators were compared for the network performance and for the time required to execute the thermographic inspection. Our tool is based on a convolutional neural network that allows to immediately recognize a failure in a PV panel reaching a very high accuracy. Considering a dataset of 1000 images that refer to different acquisition protocols, it was reached an accuracy of 99% for a convolutional neural network with 30 min of computational time on Low Mid-Range CPU. While a dataset of 200 sectioned images, the same tool achieved 90% accuracy with a multi-layer perceptron architecture and 100% accuracy for a convolutional neural network. The proposed methodology offers an open alternative and a valid tool that improves the resolution of image classification for remote failure detection problems and that can be used in any scientific sector.