Image Recognition (Computer Vision)

A research area in Computer Vision focuses on the identification of articulated objects, such as human actions and movements of the hand, which can be used in human-computer interaction, surveillance, and other tracking systems. Two problems arise: identify when two articulated objects in different stances are in the same class of objects, and differentiate the distinct positions of the same object. In both cases, it is necessary to know how correspond the different points or regions of such objects standing in different attitudes. This article presents the Contour-Point Signature; a point descriptor that allows to establish a method to achieve the better matching of points between two figures, and to thus obtain a transformation which relates them. With this descriptor, we can achieve more accurate shape features and implement more efficient retrieval under multi-resolution. In addition, CPS is robust to rigid translation, scaling, rotation and independent of the origin point. A measure of dissimilarity between two figures for classifying various human postures in a video sequence is also presented.

The increasing demand of World Wide Web raises the need of predicting the user's web page request. The most widely used approach to predict the web pages is the pattern discovery process of Web usage mining. This process involves inevitability of many techniques like Markov model, association rules and clustering. Fuzzy theory with different techniques has been introduced for the better results. Our focus is on Markov models. This paper is introducing the vague Rules with Markov models for more accuracy using the vague set theory.

The search for one digital image among the vast quantity of images is a real dilemma when it is based on image content rather than metadata such as eigenvalues or frequency coefficients. The search results for most of the existing methods are satisfactory but still included irrelevant images for the target image. We have introduced a new Image recognition method based on Wavelet Transform (DWT) and Singular Value Decomposition (SVD) that is capable of retrieving most of the images similar to the target image. Our Method used DWT to transfer the target image from the spatial domain into frequency domain in which it is divided into four sub bands, Low-Low (LL), Low-High (LH), High-Low (HL) and High-High (HH) frequencies. We have applied three levels of 2-D DWT to concentrate the image illumination components into the third-level LL sub band and then applied SVD to extract its singular values as the reliable and robust features for the recognition. We also calculated the mean and variance of the low frequency LL sub-band for the input image to create a feature vector composed of mean, variance, and singular values of the LL sub band, as well as the coefficients of the LL sub band. Once the feature vector is formed it is compared with the feature vectors stored in our database to recognize and retrieve the images that are similar to target image based on the minimum error criteria.

Illumination invariance remains the most researched, yet the most challenging aspect of automatic face recognition. In this paper we investigate the discriminative power of colour-based invariants in the presence of large illumination changes between training and test data, when appearance changes due to cast shadows and non-Lambertian effects are significant. Specifically, there are three main contributions: (i) we employ a more sophisticated photometric model of the camera and show how its parameters can be estimated, (ii) we derive several novel colour-based face invariants, and (iii) on a large database of video sequences we examine and evaluate the largest number of colour-based representations in the literature. Our results suggest that colour invariants do have a substantial discriminative power which may increase the robustness and accuracy of recognition from low resolution images.

La digitalizzazione delle iscrizioni: metodi e tecniche Poche discipline possono vantare di aver digitalizzato la quasi integrità del patrimonio documentario di loro interesse come l'epigrafia latina e greca ([12];[13]). Con trent'anni di esperienza nella digitalizzazione delle iscrizioni, l'epigrafia può vantare di essere arrivata molto vicino a questo risultato, 1 ma ha anche imparato molto sulla sostenibilità di tale impresa e sul continuo impegno che essa comporta. Nel corso del progetto EAGLE (Europeana Network for Ancient Greek and Latin Epigraphy, 2013-2016) è stato per esempio realizzato un modello, basato sull'affermato standard TEI/EpiDoc ([1];[4];[10]) in grado di garantire una mappatura a CIDOC-CRM e all'EMD (Europeana Metadata Model). Questo lavoro ha reso possibile non solo lo sviluppo del portale EAGLE con le sue funzionalità fino ad allora solo auspicate dalla comunità internazionale degli epigrafisti, ma ha anche permesso di raccogliere un significativo consorzio di partner e di produrre risorse aggiuntive, come ad esempio la Virtual Exhibition "Signs of Life" che raccoglie link a progetti, 3D, infografica in una introduzione per non addetti ai lavori ([9]).

We would like to invite you to join this exciting new project as a chapter contributor. Since this is a textbook, a great deal of this chapter entails a survey on the topic under the paradigm of cyber-physical systems, what can be done onboard and remotely, the distributed nature of the system and some exercises on futurology (anticipating trends can shed some light on upcoming designs). IET will bring great visibility to your work. You are welcome to suggest another topic/chapter title if you feel it would be more suitable. Each chapter should be around 20-25 pages each and can be submitted as a Word or Latex File. The IET will send you additional information (formatting, permission form, etc.) with the contributor's agreement once you have agreed to contribute to the book. Visit http:// www.theiet.org/resources/author-hub/books/index.cfm to get all information you need as a contributor to an IET research-level book. Each book is expected to have a total number of 500 printed pages (based on approximately 550 words per page with a 20% allowance for figures and tables). We have included a tentative schedule and list of topics below. If this is something you would consider, please send me the title of your chapter, a short description/abstract of the chapter content, and your full contact details. We will expect original content and new results for this book. You can, of course, reuse published material but the percentage of material reuse for the chapter should be less than 40%. The IET will run a piracy software on the full manuscript to control that you are including original material and will reject chapters who contain a large amount of already-published material so please do take this into consideration.

It is well known that this soiling can reduce the generation efficiency in PV system. In some case according to the literature of loss of energy production in photovoltaic systems can reach up to 50%. In the industry there are various types of cleaning robots, they can substitute the human action, reducing cleaning cost, be used in places where access is difficult, and increasing significantly the gain of the systems. In this paper we present an application of computer vision method for soiling recognition in photovoltaic modules for autonomous cleaning robots. Our method extends classic CV algorithm such Region Growing and the Hough. Additionally, we adopt a pre-processing technique based on Top Hat and Edge detection filters. We have performed a set of experiments to test and validate this method. The article concludes that the developed method can bring more intelligence to photovoltaic cleaning robots.

- SISTEM GRAFIKA KOMPUTER
Teknologi Display
Raster-scan Display
Random-scan Display 
Monitor Color CRT
Flat Panel Display
Peralatan Input Interaktif 
Peralatan Hardcopy
Perangkat Lunak Grafika 
- OUTPUT PRIMITIF
Titik dan Garis
Algoritma pembentukan garis
Algoritma garis DDA
Algoritma Garis Bressenhem
Algoritma Pembentukan Lingkaran
- ATRIBUT OUTPUT PRIMITIF
Atribut Garis
Tipe Garis
Tebal Garis
Pilihan Pen dan Brush
Warna Garis
Fill Area Primitif
Algoritma Boundary-Fill
Algoritma Flood-Fill
Pembentukan Karakter 
Antialiasing
Supersampling atau Postfiltering
Area Sampling
Pixel Phasing

International Journal of Computer Vision and machine learning (IJCVML) is an open access, peer-reviewed journal that publishes articles which contribute new results in all areas of the advanced vision computing. The goal of this journal is to bring together researchers and practitioners from academia and industry to focus on understanding advances in vision computing and establishing new collaborations in these areas. Authors are solicited to contribute to the journal by submitting articles that illustrate research results, projects, surveying works and industrial experiences that describe significant advances in the areas of vision computing.

Efficient and accurate object detection has been an important topic in the advancement of computer vision systems. With the advent of machine learning and deep learning techniques, the accuracy for object detection has increased drastically. The project aims to incorporate state-of-the-art technique for object detection with the goal of achieving high accuracy with a real-time performance. In this project, we use a completely machine learning with opencv and deep learning based approach to solve the problem of object detection in an end-to-end fashion. The network is trained on the most challenging publicly available dataset, on which a object detection challenge is conducted annually. The resulting system is fast and accurate, thus aiding those applications which require object detection.

This article proposes a new methodological approach to measure and map the size of coarse clasts on a land surface from photographs. This method is based on the use of the Mask Regional Convolutional Neural Network (R-CNN) deep learning algorithm, which allows the instance segmentation of objects after an initial training on manually labeled data. The algorithm is capable of identifying and classifying objects present in an image at the pixel scale, without human intervention, in a matter of seconds. This work demonstrates that it is possible to train the model to detect non-overlapping coarse sediments on scaled images, in order to extract their individual size and morphological characteristics with high efficiency (R 2 = 0.98; Root Mean Square Error (RMSE) = 3.9 mm). It is then possible to measure element size profiles over a sedimentary body, as it was done on the pebble beach of Etretat (Normandy, France) in order to monitor the granulometric spatial variability before and after a storm. Applied at a larger scale using Unmanned Aerial Vehicle (UAV) derived ortho-images, the method allows the accurate characterization and high-resolution mapping of the surface coarse sediment size, as it was performed on the two pebble beaches of Etretat (D50 = 5.99 cm) and Hautot-sur-Mer (D50 = 7.44 cm) (Normandy, France). Validation results show a very satisfying overall representativity (R 2 = 0.45 and 0.75; RMSE = 6.8 mm and 9.3 mm at Etretat and Hautot-sur-Mer, respectively), while the method remains fast, easy to apply and low-cost, although the method remains limited by the image resolution (objects need to be longer than 4 cm), and could still be improved in several ways, for instance by adding more manually labeled data to the training dataset, and by considering more accurate methods than the ellipse fitting for measuring the particle sizes.

The realtime hand motion recognition under unconstrained environment is a challenging computer vision problem. The change in illumination and non-uniform background condition makes it very difficult to perform real-time hand gesture recognition operations. This paper demonstrates a region-based convolutional neural network for real-time hand gesture recognition. The custom dataset is captured under unconstrained environments. The Faster region-based convolutional neural network (Faster-RCNN) with Inception V2 architecture is used to extract the features from the proposed region. The average precision, average recall, and F1-score are analyzed by training the model with a learning rate of 0.0002 for Adaptive Moment Estimation (ADAM) and Momentum optimizer, 0.004 for RMSprop optimizer. The ADAM optimization algorithm resulted in better precision, recall and F1-score values after evaluating custom test data. For ADAM optimizer with intersection over union (IoU) =0.5:0.95, the observed average precision is 0.794, average recall is 0.833, and the F1-score is 0.813. For an IoU of 0.5, ADAM optimizer resulted in 0.991 average precision with a prediction time of 137ms.

The advent of social media, particularly Twitter, raises many issues due to a misunderstanding regarding the concept of freedom of speech. One of these issues is cyberbullying, which is a critical global issue that affects both individual victims and societies. Many attempts have been introduced in the literature to intervene in, prevent, or mitigate cyberbullying; however, because these attempts rely on the victims' interactions, they are practical. Therefore, detection of cyberbullying without the involvement of the victims is necessary. In this study, we attempted to explore this issue by compiling a global dataset of 37,373 unique tweets from Twitter. Moreover, seven machine learning classifiers were used, namely, Logistic Regression (LR), Light Gradient Boosting Machine (LGBM), Stochastic Gradient Descent (SGD), Random Forest (RF), AdaBoost (ADB), Naive Bayes (NB), and Support Vector Machine (SVM). Each of these algorithms was evaluated using accuracy, precision, recall, and F1 score as the performance metrics to determine the classifiers' recognition rates applied to the global dataset. The experimental results show the superiority of LR, which achieved a median accuracy of around 90.57%. Among the classifiers, logistic regression achieved the best F1 score (0.928), SGD achieved the best precision (0.968), and SVM achieved the best recall (1.00).

Biometric recognition systems have advanced significantly in the last decade and their use in specific applications will increase in the near future. The ability to conduct meaningful comparisons and assessments will be crucial to successful deployment and increasing biometric adoption. The best modality used as unimodal biometric systems are unable to fully address the problem of higher recognition rate. Multimodal biometric systems are able to mitigate some of the limitations encountered in unimodal biometric systems, such as non-universality, distinctiveness, non-acceptability, noisy sensor data, spoof attacks, and performance. More reliable recognition accuracy and performance are achievable as different modalities were being combined together and different algorithms or techniques were being used. The work presented in this paper focuses on a bimodal biometric system using face and fingerprint. An image enhancement technique (histogram equalization) is used to enhance the face ...

IBM Visual Recognition utiliza algoritmos de deep learning para analizar imágenes de escenas, objetos, texto, entre otros elementos.

Realizar hallazgos y construir una narrativa adecuada para comunicarlos son habilidades fundamentales en Ciencia de Datos. El principal ingrediente es la curiosidad.

The thesis concentrates on computational methods pertaining to ancient ostraca - ink on clay inscriptions, written in Hebrew. These texts originate from the biblical kingdoms of Israel and Judah, and dated to the late First Temple period (8th – early 6th centuries BCE). The ostraca are almost the sole remaining epigraphic evidence from the First Temple period and are therefore important for archaeological, historical, linguistic, and religious studies of this era. This “noisy” material offers a fertile ground for the development of various “robust” image analysis, image processing, computer vision and machine learning methods, dealing with the challenging domain of ancient documents’ analysis. The common procedures of modern epigraphers involve manual and labor-intensive steps, facing the risk of unintentionally mixing documentation with interpretation. Therefore, the main goal of this study is establishing a computerized paleographic framework for handling First Temple period epigraphic material. The major research questions, addressed in this thesis are: quality evaluation of manual facsimiles; quality evaluation of ostraca images; automatic binarization of the documents and its subsequent refinement; quality evaluation of binarizations on global and local levels; identification of different writers between inscriptions (two distinct methods are proposed); image segmentation (with improvements over the classical Chan-Vese algorithm); and letters’ shape prior estimation. The developed methods were tested on real-world archaeological and modern data and their results are found to be favorable.

This study analyzes a low-cost reliable real-time optimal monitoring platform for fused filament fabrication-based open source 3-D printing. An algorithm for reconstructing 3-D images from overlapping 2-D intensity measurements with relaxed camera positioning requirements is compared with a single camera solution for single side 3-D printing monitoring. The algorithms are tested for different 3-D object geometry and filament colors. The results showed that both of the algorithms with a single and double camera system were effective at detecting a clogged nozzle, incomplete project, or loss of filament for a wide range of 3-D object geometries and filament colors. The combined approach was the most effective and achieves 100 percent detection rate for failures. The combined method analyzed here has a better detection rate and a lower cost compared to previous methods. In addition, this method is generalizable to a wide range of 3-D printer geometries, which enables further deployment of desktop 3-D printing as wasted print time and filament are reduced, thereby improving the economic advantages of distributed manufacturing.

Logistic Regression (Preloaded Dataset) scikit-learn comes with a few small datasets that do not require to download any file from some external website. The digits dataset we will use is one of these small standard datasets. These datasets are useful to quickly illustrate the behavior of the various algorithms implemented in the scikit. They are however often too small to be representative of real world machine learning tasks. After learning the basics of logisitic regression, we will use the MNIST Handwritten digit database Each datapoint is a 8x8 image of a digit.

youtube video: https://www.youtube.com/watch?v=71iXeuKFcQM
Logistic Regression Jupyter Notebook (Digits Dataset): https://github.com/mGalarnyk/Python_Tutorials/blob/master/Sklearn/Logistic_Regression/LogisticRegression_toy_digits.ipynb
Logistic Regression Jupyter Notebook (MNIST Dataset): https://github.com/mGalarnyk/Python_Tutorials/blob/master/Sklearn/Logistic_Regression/LogisticRegression_MNIST.ipynb

As technology continues to evolve, some traditional practices tend to fade. Stenography is a practice of writing dictations in terms of speed devised by Isaac Pitman. The said practice had been around the early 19th century to address the slow phased transcription process in court trials. However, such a method can diminish as people tend to rely moreover upon recently developed advanced technologies. This study aims to use Convolutional Neural Networks and image processing techniques to preserve stenography by allowing machines to recognize and use the said writing practice in modern times. This research used 2000 common court stenography words and phrases as the core dataset for training, validation, and testing. Application of canny edge detection provides the detection of edges to provide a better classification of stenographic writings. Results generated show that through Convolutional Neural Networks, stenography writings can be recognized even by the machines. Canny Edge Detection highly supported the improvement of results by allowing the model to focus more on the necessary features for better classification accuracy. The researchers also determined future enhancements and works that can contribute in the given research.

With more and more attentions paid on health, people begin to care about healthy diet options created by experts on nutrition. However, it will take a long time to observe the effects by taking healthy diet. This causes great difficulty for users to follow the healthy diet strictly. Most existing applications are not user-friendly in inputting information to the application. Then it becomes difficulty to track for exact health status. This paper proposes an android application which can be trained to recognize different kinds of food and facilitate the information input through phone camera using machine learning algorithms. Thus, nutritional information can be fed in application accurately.

The paper explains the development of a classification program which is to be integrated to a robot that will autonomously play chess. The problem is to perform a classification on a 12 class data set of chess pieces which works on a real-time video feed. Various approaches have been discussed for the problem; finally Convolutional Neural
Network was chosen to develop the solution.

The paper develops two different approaches to solve the problem. One approach uses a single Convolutional Neural Network for classification. The other divides the problem into two classification task, classifying of the chess piece color and classifying the chess piece based on shape. The models developed from both the approaches are compared based on accuracy and computational time.

Cross validation was used to ensure the classification accuracy results are repeatable. To understand the accuracy in classifying each class, a confusion matrix was generated and the important features were visualized to understand the learnt features. The computation time for all the function calls on the program were measured and the average was calculated. Finally, the robustness of the 1-Step model to the change in object size was experimentally determined.

Deep learning and image processing are two areas of great interest to academics and industry professionals alike. The areas of application of these two disciplines range widely, encompassing fields such as medicine, robotics, and security and surveillance. The aim of this book, 'Deep Learning for Image Processing Applications', is to offer concepts from these two areas in the same platform, and the book brings together the shared ideas of professionals from academia and research about problems and solutions relating to the multifaceted aspects of the two disciplines. The first chapter provides an introduction to deep learning, and serves as the basis for much of what follows in the subsequent chapters, which cover subjects including: the application of deep neural networks for image classification; hand gesture recognition in robotics; deep learning techniques for image retrieval; disease detection using deep learning techniques; and the comparative analysis of deep data and big data. The book will be of interest to all those whose work involves the use of deep learning and image processing techniques

This undergraduate thesis describes  a voice recognition system used for home appliances control.  Here it is tested using 3 loads but more loads can be added. Appliances can be controlled both using the voice and using ordinary push buttons found at a control post.

—The paper presents an automated system for classification of fruits. A dataset containing five different fruits was constructed using an ordinary camera. All the fruits were analyzed on the basis of their color (RGB space), shape and texture and then classified using different classifiers to find the classifier that gives the best accuracy. Gray Level Co-occurance Matrix(GLCM) is used to calculate texture features .Best accuracy was achieved by support vector machine (svm). All the processing was carried out in Matlab.

12th International Conference on Artificial Intelligence, Soft Computing and Applications (AIAA 2022) will provide an excellent international forum for sharing knowledge and results in theory, methodology and applications of Artificial Intelligence, Soft Computing. The Conference looks for significant contributions to all major fields of the Artificial Intelligence, Soft Computing in theoretical and practical aspects. The aim of the
Conference is to provide a platform to the researchers and practitioners from both academia as well as industry to meet and share cutting-edge development in the field.

Identifying spatio-temporal synchrony in a complex, interacting and oscillatory coupled-system is a challenge. In particular, the characterization of statistical relationships between environmental or biophysical variables with the multivariate data of pandemic is a difficult process because of the intrinsic variability and non-stationary nature of the time-series in space and time. This paper presents a methodology to address these issues by examining the bivariate relationship between Covid-19 and temperature time-series in the time-localized frequency domain by using Singular Value Decomposition (SVD) and continuous cross-wavelet analysis. First, the dominant spatio-temporal trends are derived by using the eigen decomposition of SVD. The Covid-19 incidence data and the temperature data of the corresponding period are transformed into significant eigen-state vectors for each spatial unit. The Morlet Wavelet transformation is performed to analyse and compare the frequency structure of the dominant trends derived by the SVD. The result provides cross-wavelet transform and wavelet coherence measures in the ranges of time period for the corresponding spatial units. Additionally, wavelet power spectrum and paired wavelet coherence statistics and phase difference are estimated. The result suggests statistically significant coherency at various frequencies providing insight into spatio-temporal dynamics. Moreover, it provides information about the complex conjugate dynamic relationships in terms phases and phase differences.

A huge amount of paper based documents with valuable information is available in churches, libraries, caves, governmental and private institutions in a printed, typewritten and handwritten format. To enable those documents accessible and searchable, Optical Character Recognition (OCR) systems play a vital role by converting them into their digital format. Some researchers attempted to develop Amharic OCR systems. However, OCR systems are not yet applicable for real life document images that contain column blocks, graphics, tables, lines, logos and other shapes. Moreover, the effectiveness of the system is highly dependent on the text segmentation output. This study attempts to explore an effective page and text segmentation method to improve the applicability and performance of Amharic OCR for real life documents. Accordingly, a skew correction and page segmentation algorithms based on Hough Transform, Morphological Dilation, and Connected Component (CC) Analysis are tested, and 90.47%, 92.31%, 96.67% and 71.43% accuracy is obtained for detecting tables, graphics, column blocks and titles individually. Three noise filtering and two binarization techniques are tested and wiener coupled sauvola found to perform best. Text segmentation methods based on projection profile, morphological dilation and CC Analysis are experimented on four noise levels (i.e. low, medium, high and very-high) documents. Projection profile coupled vertical dilation performs best by scoring 100% accuracy to segment text lines in low and medium noise levels. An image smoothing based method is proposed and 99.18% accuracy is registered to extract lines from inkbleeded documents. Vertical projection profile method is applied to extract words and 99.23%, 96.26%, 87.12% and 54.80% accuracy is registered for each noise levels respectively. A new method based on CC Analysis is introduced to segment overlapping characters, and besides to detect and split connected characters. An accuracy of 87.61% and 82.29% is obtained for low and medium noise levels and 50.64% for high and very high noise levels. By integrating it with the Amharic OCR system, recognition accuracy rate of 79.13% and 59.07% are registered for the proposed and vertical projection profile method respectively, which is a promising result. However, since the developed character segmentation technique fails to segment characters with discontinuity, and detects long characters as connected character for real life documents, there is a need to explore noise tolerant segmentation methods.

Scope of the book:
This book focusses on the technical concepts of deep learning and its associated branch Neural Networks for the various dimensions of image processing applications. The proposed volume intends to bring together researchers to report the latest results or progress in the development of the above-mentioned areas. Since there is a deficit of books on this specific subject matter, the editors aim to provide a common platform for researchers working in this area to exhibit their novel findings.
Topics of Interest:
This book solicits contributions, which include the fundamentals in the field of Deep Artificial Neural Networks and Image Processing supported by case studies and practical examples. Each chapter is expected to be self-contained and to cover an in-depth analysis of real life applications of neural networks to image analysis.

We created a robust computer vision classifier that performs well in a dataset that contains perturbations. To achieve this, we used various data augmentation and other deep learning model techniques, such as model ensembling, image denoising, adversarial training, and attention networks. These methods helped improve the robustness against both naturally perturbed and adversarial datasets. In addition, we created class action map visualizations for our models to help understand how the model makes its classification decisions. Through these techniques, we achieved a 71.1% Top 1 Accuracy and 90.0% Top 5 Accuracy on the Tiny-ImageNet classification challenge.

Augmented reality have undergone considerable improvement in past years. Many special techniques and hardware devices were developed, but the crucial breakthrough came with the spread of intelligent mobile phones. This enabled mass spread of augmented reality applications. However mobile devices have limited hardware capabilities, which narrows down the methods usable for scene analysis. In this article we propose an augmented reality application which is using cloud computing to enable using of more complex computational methods such as neural networks. Our goal is to create an affordable augmented reality application suitable which will help car designers in by 'virtualizing' car modifications.

Diabetes affects a lot of people everywhere, and has associated complications such as vision loss, heart failure and stroke. Among a group of 100 patients with diabetes, 10 people would likely have diabetes-related eye problems. An ophthalmologist (eye doctor) would have to check the eyes of all 100 patients periodically to find out who had problems. The proposed system tend to analysis the retina fundus images for early detect of any sign related to eyes problems due to diabetic. The image analysis system is broken down into a number of modules which sequentially process the input image to extract information at different levels. These modules included detection the exudates and hemorrhage, blood vessel touristy measure, blood vessel diameter, and each of these modules may need another steps for analyzing of image. The results were highly promised and the percentage of early detection for normal and abnormal retina was 100%.

Countless learning tasks require dealing with sequential data. Image captioning, speech synthesis, and music generation all require that a model produce outputs that are sequences. In other domains, such as time series prediction, video analysis, and musical information retrieval, a model must learn from inputs that are sequences. Interactive tasks, such as translat- ing natural language, engaging in dialogue, and controlling a robot, often demand both capabilities. Recurrent neural networks (RNNs) are connec- tionist models that capture the dynamics of sequences via cycles in the network of nodes. Unlike standard feedforward neural networks, recurrent networks retain a state that can represent information from an arbitrarily long context window. Although recurrent neural networks have tradition- ally been difficult to train, and often contain millions of parameters, recent advances in network architectures, optimization techniques, and paral- lel computation have enabled successful large-scale learning with them. In recent years, systems based on long short-term memory (LSTM) and bidirectional (BRNN) architectures have demonstrated ground-breaking performance on tasks as varied as image captioning, language translation, and handwriting recognition. In this survey, we review and synthesize the research that over the past three decades first yielded and then made practical these powerful learning models. When appropriate, we reconcile conflicting notation and nomenclature. Our goal is to provide a self- contained explication of the state of the art together with a historical perspective and references to primary research.

Convolutional neural network (CNN) is one of the most prominent architectures and algorithm in Deep Learning. It shows a remarkable improvement in the recognition and classification of objects. This method has also been proven to be very effective in a variety of computer vision and machine learning. As in other deep learning, however, training this approach is interesting yet challenging. Recently, some metaheuristic algorithms have been used to optimize CNN using Genetic Algorithm, Particle Swarm Optimization, Simulated Annealing and Harmony Search. In this paper, another type of metaheuristic algorithms with different strategy has been proposed, i.e. Microcanonical Annealing to optimize Convolutional Neural Network. The performance of the proposed method is tested using the MNIST and CIFAR-10 datasets. Although experiment results of MNIST dataset indicate the increase in computation time (1.02x-1.38x), nevertheless this proposed method can considerably enhance the performance of the original CNN (up to 4.60%). On the CIFAR10 dataset, currently, state of the art is 96.53% using fractional pooling, while this proposed method achieves 99.14%.