OCR and text recognition

Distortion is the moment at which the physical means of transmitting a text irrupt into a reader's experience of it. I will discuss distortion here as a phenomenon occurring in printed materials, but I do not wish to exclude other recognisable instances such as static in a radio signal, or white noise on a cathode-ray-tube television screen. To speak of distortion as an impairment, however, is to confirm the Platonic priority of another book-object. As textual scholars have shown, locating this authoritative ur-text can be, to say the least, difficult. Distortion is not a condition befalling individual textual artefacts (‘book-objects’) here and there: there is no non-distorted reading. In fact, a belief in non-distorted reading amounts to believing in pure, transparent mediation that leaves no trace of itself on the mediated. Leaving aside for the purposes of this essay the profound political implications of placing the labour of mediation under erasure, I will offer here a heuristic of distortion. Absent such a heuristic, ‘distorted’ book-objects and their texts will always be treated as degraded derivatives. The corollary of that treatment is to elevate an unknown, possibly non-existent ur-text to a position of unimpeachable authority, and to degrade both the experience of reading and of the embodied text and reader. We can solve the problem of distortion by placing the irreducible phenomenological singularity of the book-object at the centre of our experience of a text. This is existential reading: you are reading what you are reading and not anything else.

This research focused on the methods of optical character recognition technologies that applied to android-based translator application. The high number of deaf-born babies in Indonesia causes a high number of children who will grow but cannot recognize various types of letters yet because they are limited by their physical abilities. So, to help introduce letters more broadly, effectively, and efficiently, the latest technological innovations are needed, one of which is the application of converting images containing text into text and translating into sign language. In making this android-based application used Optical Character Recognition (OCR) technology by using Tesseract as an engine of OCR (Optical Character Recognition) for character or letter recognition.

Artificial neural networks are information processing systems that have characteristics similar to human neural networks. Learning models need to be done on an artificial neural network before being used to solve problems by examining and correcting any errors that occur during the learning process. Developments in information technology affect particular scientific expertise. Field of image processing does not become one the only method of solving a problem, but the current image processing combined with artificial intelligence to examine or look for a solution in a variety of applications. Akshara Java is one of the priceless cultural heritages. Form of script and art-making becomes a relic that deserves to be preserved. Not only in Java, but Akshara Java is also used in the Sunda and Bali, although there is little difference in the writing used the same script. The purpose of this research is to create a model of image processing and converted into text into voice so that people ...

In the last decade, there has been international interest in providing special e-services for our social partnership that are visually disabled. In particular, they face hard and numerous problems in their daily activities such as paying for products or service's fee or bills electronic-payment. In this paper, a proposed e-service is introduced to visually impair in order to facilitate the process of e-payment. This e-service is used to allow them to pay their bills without external help from humans with easy and simple steps. The proposed e-service depends on two technical features for reading and listening to their bills electronically such as (Optical Character Recognition) OCR and (Text-to-Speech) TTS thus, after converting the snapshot image into (digital numbers) or money value from. Furthermore, it is directly connected to their bank accounts by using Near Field Communication (NFC) technology that is available on their smartphones. Therefore, the payment is done with high privacy and short-range communication. Therefore, TTS technique will be used to allow visually impaired to listen. Moreover, a "PAY" voice command is used to give the payment process in order to confirm and approve the current payment transaction. A large sector of blind disabilities will get more benefits after applying this integrated e-service. It allows them to secure pay their purchases at any time, as well as quickly as possible and will be a convenience payment method.

Optical character recognition for the English text may be considered one of the most important research topics, whether, printed or handwritten. Although excellent results have been reached in the English text, there is a lack of this type of research in the Arabic text. This is because of the nature of the Arabic alphabet, and the multiplicity of forms of the same letter. Arabic handwritten character recognition (AHCR) systems involve several issues, and challenges from finding a suitable, and public Arabic handwritten text dataset phase to recognition, and classification phase passing through seg-mentation, and feature extraction phases. The paper objectives are: Firstly, a large, and complex Arabic handwritten characters' dataset (HMBD) is presented for training, testing, and validation phases, as well as, discussing its collection, preparation, cleaning, and preprocessing. Secondly, we introduce a deep learning (DL) system with two convolutional neural network (CNN) architectures (named HMB1 and HMB2); with the appliance of optimization, regularization, and dropout techniques. This system can serve as a baseline for future research on handwritten Arabic text. Different performance metrics were calculated such as accuracy, recall, precision, and F1. 16 experiments were applied to the described system using HMBD, and another two datasets: CMATER, and AIA9k. Experiments' results were captured and compared to study the effects of weight initializers, optimizers, data augmentation, and regularization on overfitting, and accuracy. He Uniform weight initializer and AdaDelta optimizer reported the highest accuracies. Data augmentation showed an improvement in the accuracies. HMB1 reported testing accuracy of 98.4% with 865,840 records using augmentation on HMBD. CMATER and AIA9k datasets were used for validating the generalization. Data augmentation was applied, and the best results were 100%, and 99.0% for testing accuracies, respectively. A cross-over validation between the described architectures, and a previous state-of-the-art architecture, and dataset was performed in two phases. First, the previous control architecture cannot generalize for the presented dataset in the current study. Second, the study described archi-tectures generalize for the control dataset, with higher accuracies (97.3%, and 96.8% for HMB1, and HMB2, respectively), than the reported accuracy in the selected control study.

BanglaOCR is currently the only open source optical character recognition (OCR) software for the Bangla (Bengali) script developed by the Center for Research on Bangla Language Processing (CRBLP). Tesseract, maintained by Google, is considered
to be one of the most accurate free open source OCR engines currently available. In this paper, we present a new OCR for the Bangla/Bengali script that combines the recognition power of Tesseract and the Bangla script processing power of BanglaOCR
by integrating the Tesseract recognition engine into BanglaOCR. We first present the complete methodology to build the combined OCR, followed by the implementation strategy (in python programming language). In this project, we focus on the training
data preparation process, Tesseract integration procedure and the post-processing techniques. The techniques described in this paper can be readily applied to build OCRs for other scripts as well.

A few decades ago, research in the field of Character Recognition was limited to document images acquired with flatbed desktop scanners. The usability of such systems is limited as they are not portable because of large size of the scanners and takes a lot of time to scan an image. But, with the advancement of processing speed, internal memory and built-in digital cameras of hand-held mobile devices, a new trend of research has emerged into picture. We may think of running OCR applications on such devices for having real time imaging results. This Paper summarise some commercially available applications to recognise text from documents. Most of the available systems work for European scripts which are based on Roman alphabets.

The paper is a comprehensive review of the current research trends in the area of Arabic language especially state-of-the-art approaches to highlight the current status of diverse research aspects of that area to facilitate the adaption and extension of previous systems into new applications and systems. The Arabic language has deep, widespread and unexplored scope to research although the tremendous effort and researches that had been done previously. Modern state-of-the-art methods and approaches with fewer errors are required according to the high speed of hardware and technology development. The focus of this article will be on the offline Arabic handwritten text recognition as it is one of the most important topics in the Arabic scope. The main objective of this paper is critically analyzing the current researches to identify the problem areas and challenges faced by the previous researchers. This identification is intended to provide many recommendations for future advances in the area. It also compares and contrasts technical challenges, methods and the performances of handwritten text recognition previous researches works. It summarizes the critical problems and enumerates issues that should be considered when addressing these tasks. It also shows some of the Arabic datasets that can be used as inputs and benchmarks for training, testing and comparisons. Finally, it provides a fundamental comparison and discussion of some of the remaining open problems and trends in that field.

Mass-digitization projects in libraries have received world-wide attention. Today a significant amount of historical books, newspapers and journals are already digitized and available online for search and retrieval. The situation is totally different if we look at archives. Though major state and regional archives contain tens or even hundreds of shelf kilometers with most often unique material, the state of digitization lags far behind the library world. One reason may be the absence of a technology that can deal with handwritten documents and convert them into editable text. This paper introduces the concept of a comprehensive Transcription and Recognition Platform (TRP) which may support the development of Handwritten Text Recognition (HTR) technology, based on the assumption that this could best be done within a cooperative model where archives, humanities scholars, computer scientists and the general public are involved and provide specific contributions according to their individual mission and interest.

The purpose of this project is to develop an Automatic gate control application which recognizes license plates from cars at entrance gate and take an action to let cars enter or not.  And regular PC with camera, catches image frame when a car is detected, the camera takes a picture of the car. The picture is then sent to the MATLAB to filter unwanted noise (illumination, blur) etc. leaving the characters of the number plate. The characters can then be identified using algorithm to recognition the characters With number plates coming in various shapes, sizes, fonts and colors, the algorithms needs to be refined to increase its efficiency. Most of the research on system targets hardware that uses programming by (MATLAB) and  processing such as microprocessors and microcontroller in our project we uses components to interface  them together in  our project  (MATLAB, Arduino , camera, mechanical design,).

Optical character recognition is the machine replication of hu-
man reading . It can be described as Mechanical or electronic
conversion of scanned images where images can be hand written, type-
written or printed text. This paper presents Google’s open source
Optical Character Recognition software Tesseract. We will give an
overview of the algorithms used in the various stages in the pipeline
of Tesseract. In particular we will focus on the aspects that are novel
or at least unusual in Tesseract compared to other OCR engines.

Optical character recognition has remained a challenge for comics, given the high variability of placement of text on the page, the wide variety of frequently handwritten fonts, and the limited availability and small size of datasets. This paper reports on currently ongoing work on an OCR pipeline that includes text spotting with the help of a U-Net based fully convolutional neural network and OCR training with the open-source software Calamari, which was performed on the "Graphic Narrative Corpus" of book-length graphic novels written in English. Based on the results of the OCR training, we then present an analysis of the textual properties of 129 graphic novels correlated with page length, historical development, and genre affiliation.

This paper was written to create awareness among the library
professionals about AIDC (Automatic Identification and Data Capture)
technologies and their potential application in libraries/information
centres. Data entry is an inseparable part of the library operations. This
is an repetitive job and slows down the speed of the system as a whole.
So, AIDC technologies have been described in this paper as an
probable solution. Each technology has been introduced individually
and its working has been described. The technologies which have been
included are, bar code, RFID, biometrics, OCR, MICR, magnetic strips,
smart card and voice recognition. After that the applicability of each
individual technology in libraries has been described briefly. Towards
the conclusion of the paper the future of AIDCs have been discussed in
view of the changing scenario with the introduction of born digital
documents and internet.

 Abstract: In current situation, we come across various problems in traffic regulations in India which can be solved with different ideas. Riding motorcycle/mopeds without wearing helmet is a traffic violation which has resulted in increase in number of accidents and deaths in India. Existing system monitors the traffic violations primarily through CCTV recordings, where the traffic police have to look into the frame where the traffic violation is happening, zoom into the license plate in case rider is not wearing helmet. But this requires lot of manpower and time as the traffic violations frequently and the number of people using motorcycles is increasing day-by-day. What if there is a system, which would automatically look for traffic violation of not wearing helmet while riding motorcycle/moped and if so, would automatically extract the vehicles' license plate number. Recent research have successfully done this work based on CNN, R-CNN, LBP, HoG, HaaR features,etc. But these works are limited with respect to efficiency, accuracy or the speed with which object detection and classification is done. In this research work, a Non-Helmet Rider detection system is built which attempts to satisfy the automation of detecting the traffic violation of not wearing helmet and extracting the vehicles' license plate number. The main principle involved is Object Detection using Deep Learning at three levels. The objects detected are person, motorcycle/moped at first level using YOLOv2, helmet at second level using YOLOv3, License plate at the last level using YOLOv2. Then the license plate registration number is extracted using OCR (Optical Character Recognition). All these techniques are subjected to predefined conditions and constraints, especially the license plate number extraction part. Since, this work takes video as its input, the speed of execution is crucial. We have used above said methodologies to build a holistic system for both helmet detection and license plate number extraction.

Optical character recognition is popular field for researchers during last decade of research, which is able to successfully recognize the scanned English image into editable text form. However, optical character systems for other regional languages such as Urdu, Arabic, and Sindhi, still presents a huge challenge and implementation problems. Thus, in this paper various techniques of optical character recognition system for such low level regional languages have been discussed and analyzed. This survey paper consolidates all such techniques and presents an overview to aid researcher understand the methodology of performing and implementing OCR system for Sindhi language.

In this research we propose a statistical method and morpho-lexical analysis for correcting Arabic words as a post processor for Arabic words output from OCR systems. Dictionaries of words were built for the comparison to the attached word. The present research uses multiple knowledge sources and basing on the Arabic language properties, statistical method, morpho-lexical analysis and dictionary look-up for error detection and correction. Correction of errors in this research depends on the type of possible error, which can be: transposing two adjacent letters, rejection, replacing an incorrect letter, inserting a missing letter, substitution errors, which are most frequently committed by the OCR systems. ‫ﺍﻟ‬ ‫ﻤ‬ ‫ﺴﺘ‬ ‫ﺨ‬ ‫ﻠ‬ ‫ﺹ‬ ‫ﻴﻘﺘﺭ‬ ‫ﺇﺤﺼﺎﺌﻴﺔ‬ ‫ﻁﺭﻴﻘﺔ‬ ‫ﺍﻟﺒﺤﺙ‬ ‫ﻫﺫﺍ‬ ‫ﺡ‬ ‫ﻭﺘﺤﻠﻴل‬ ‫ﻤﻭﺭﻓﻭﻟﻭﺠﻲ‬ ‫ﺍﻟﻌﺭﺒﻴـﺔ‬ ‫ﺍﻟﻜﻠﻤـﺎﺕ‬ ‫ﻟﺘﺼﺤﻴﺢ‬ ‫ﺍﻟﺨﺎﻁﺌﺔ‬ ‫ﺒﻭﺼﻔﻬﺎ‬ ‫ﻭﺴﻴﻠﺔ‬ ‫ﺍﻟـﻀﻭﺌﻲ‬ ‫ﺍﻟﺘﻤﻴﻴـﺯ‬ ‫ﺃﻨﻅﻤـﺔ‬ ‫ﻤﻥ‬ ‫ﺍﻟﻨﺎﺘﺠﺔ‬ ‫ﺍﻟﻌﺭﺒﻴﺔ‬ ‫ﻟﻠﻜﻠﻤﺎﺕ‬ ‫ﻨﻬﺎﺌﻴﺔ‬ ‫ﻤﻌﺎﻟﺠﺔ‬ OCR. ‫ﻴ‬ ‫ﺒﺎﻟﺒﺤﺙ‬ ‫ﻠﺤﻕ‬ ‫ﻗﻭﺍﻤﻴﺱ‬ ‫ﺒﺎ‬ ‫ﻤﻊ‬ ‫ﺍﻟﻤﻌﺎﻟﺠﺔ‬ ‫ﺘﺤﺕ‬ ‫ﺍﻟﻜﻠﻤﺔ‬ ‫ﻤﻘﺎﺭﻨﺔ‬ ‫ﺒﻬﺩﻑ‬ ‫ﻟﻜﻠﻤﺎﺕ‬ ‫ﻤﻁﺎﺒﻘﺘﻬﺎ‬ ‫ﻓـﻲ‬ ‫ﺍﻟﻘﺎﻤﻭﺱ‬ .

UNIVERSIDAD DE TURKU Instituto de Lenguas y Traducción / Facultad de Humanidades

SAMMALVUO, TAPANI: Trabajo terminológico basado en corpus: caso práctico de terminología del ajedrez Trabajo de fin de máster, 74 pág., 7 pág. de apéndices Departamento de Español, Traducción de español Noviembre 2016 – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – – 

El presente trabajo de fin de máster pertenece a los campos de Terminología y Lingüística de corpus y tiene tres objetivos principales: investigar los efectos de omitir la corrección manual de un corpus escaneado en el análisis de frecuencia de palabras, proponer un método completo de extracción de términos a partir de un corpus de texto escaneado y, presentar una terminología del ajedrez de 100 términos más comunes en mi corpus electrónico.

El trabajo se divide en parte teórica y la parte empírica. La parte teórica inicia con una breve presentación del proceso OCR y los problemas que puede encontrarse en él, y luego se revisa el estado de teorías de Lingüística de corpus y de Terminología, siempre y cuando estén relacionadas con la metodología aplicada en la parte empírica.

La parte empírica presenta el caso práctico en el cual se construye un corpus electrónico del ajedrez con más de 100 millones de palabras. Luego se lo evalúa y manipula con programas informáticos adecuados, obteniendo como resultado una lista de términos candidatos. 

En la fase final yo, sirviendo como experto en la materia, elijo los 100 términos del ajedrez más frecuentes –más nueve casos de homonimia– y los analizo con los métodos terminográficos clásicos, siempre proponiendo los equivalentes asimismo en finés e inglés.

Palabras clave: terminology, corpus linguistics, Spanish language, Finnish language, English language

— Meter reading is an Application designed to automatically collecting consumption, diagnostic, and status data from utility meters and transferring the retrieved data to a central database for billing, troubleshooting, and analyzing etc. Tracking the real-time consumption on a day to day basis in the most cost effective way to identify energy wastage coupled with analysis.

The paper discusses the automatic text recognition capabilities of neural network models specifically trained to recognize different styles of Church Slavonic handwriting within the software platform Transkribus. Computed character error rates of the models are in the range of 3 to 4 percent; real-life performance shows that specifically trained models, by and large, recognize simple (non-superscript) characters correctly most of the time. Error rate is higher with superscript letters, abbreviations, and word separation. Combined models consisting of training data from different sources are capable of transcribing different styles of Slavic handwriting with low error rate. Automatic text recognition using Transkribus and the models presented in this paper can help improve the efficiency of the process of digitizing Church Slavonic manuscripts and, thus, boost the number of digitized sources available in the future.

This issue (DOI: 10.13140/RG.2.2.13132.92807) includes the following articles; P1151705543, author="Tieling Chen", title="Detail Preserving Sorted Difference Filter", P1151738585, author="Vladimir A. Kulyukin", title="GreedyHaarSpiker: An Algorithm for In Situ Detection of Highway Lane Boundaries with 1D Haar Wavelet Spikes", P1151714561, author="Poorva Waingankar and Sangeeta Joshi", title="Video Compression using Efficient Encoding Techniques for Low Bit Rate Applications", P1151733581, author="Ayush Purohit and Shardul Singh Chauhan", title="A Precise Technique for Hand Gesture Recognition", P1151748594, author="D. Boopathy and M. Sundaresan", title="Securing Images on Cloud using Multidimensional Approach", P1151747592, author="M. Anyayahan and M. Balinas and A. La Madrid and M. Laurel and C. Lopez and R. Tolentino", title="Rotational invariant Real Time Text Recognition",

Studies on handwriting recognition systems have gain a great attention since it has been considered as an important technology in computer science, especially that handwriting documents have continued to be the most used mean of communication and restoring information in human life. In this paper, a method for detecting the prescribed medications in physicians' handwritten prescriptions is presented. The K-nearest neighbor classifier is used to classify the medications in this study. Furthermore, the Gabor filter is used for feature extraction. To gain a better results, a preprocessing step is considered to improve the quality of extracted features. The simulation results show about 80 percent correct recognition rate on the test samples.

A well designed image database is very necessary for the recognition of any language and many of language of the world have their own database for the text recognition. In this paper we are presenting the comprehensive database for Sindhi Language which is highly demanding language of the Middle East countries. The database will consist of the many words, which are written by the many writers. This is novel approach for the creating and testing of Sindhi Text. This database contains the isolated characters as well as the text of Sindhi language. The database is consisting of the images of the handwritten text, ground truth information, boxes for the users writing and lines for the text recognition. Tools for the segmentation and creation of the database have been developed in MATLAB, which helps to extract the handwritten data from the designed forms. We developed the different forms for the collection of data from the different users. This database can be used by the other language which possesses the same characteristics of Sindhi Language.

There are two most popular writing styles of Urdu i.e. Naskh and Nastaliq. Considering Arabic OCR research, ample amount of work has been done on Naskh writing style; focusing on Urdu, which uses Arabic character set commonly used Nastaliq writing style. Due to Nastaliq writing style, Urdu OCR poses many distinct challenges like compactness, diagonal orientation and context character shape sensitivity etc., for OCR system to correctly recognize the Urdu text image. Due to compactness and slanting nature of Nastaliq writing style, existing methods for Naskh style would not give desirable results. Therefore, in this paper, we are presenting ligature based segmentation OCR system for Urdu Nastaliq script. We have discussed in detail various unique challenges for the Urdu OCR and different feature extraction techniques for Ligature recognition using SVM and kNN classifier. The system is trained to recognize 11,000 Urdu ligatures. We have achieved overall 90.29% accuracy tested on Urdu text images.

The purpose of this research is to improve the
recognition rate of on-line Arabic handwriting recognition
using HMM (Hidden Markov Model). Delayed strokes are
removed from the on-line Arabic word to avoid the difficulty
and the confusion caused by the delayed strokes in the
recognition process.  Dictionaries for all the words in the
ADAB database have been constructed with and without the
delayed strokes.  Word matching in both dictionaries along
with effective on-line features and careful choice of the HMM
parameters have significantly improved the recognition rate of
the proposed system over other HMM-based on-line Arabic
handwriting recognition systems.

Da molti anni ormai mezzi di stampa e addetti ai lavori sottolineano le alterne vicende del libro cartaceo e del suo equivalente digitale, l’e-book. Oltre che per le consuete, scoraggianti statistiche sull’abitudine alla lettura, gli italiani si sono distinti per un certo tradizionalismo, continuando a preferire l’esperienza più familiare, quella “fisica” delle pagine da sfogliare.  Al di là dei numeri, appare evidente che la progressiva scomparsa del libro cartaceo tradizionale spesso preconizzata dai più entusiastici fautori dell’innovazione digitale non ha avuto luogo. L’impressione è che i due canali si siano via via differenziati, in base ai contenuti, al pubblico e alle modalità di accesso e fruizione. Questo capitolo intende riflettere sulla diversa esperienza di lettura offerta dai due “media” attraverso una sommaria panoramica delle relative destinazioni d’uso.

The overall objective of READ is to implement a Virtual Research Environment where archivists, humanities scholars, computer scientists and volunteers are collaborating with the ultimate goal of boosting research, innovation, development and usage of cutting edge technology for the automated recognition, transcription, indexing and enrichment of handwritten archival documents. This Virtual Research Environment will gather thirteen research organisations and archives and make the technology available to the public in various forms: Open Access, Open Research Data, Open Source. A protoype implementation can be explored at the Transkribus website.

This paper proposes a method to detect and identify the vehicle number plate that will help in the linking it with owner's bank account or a pre paid account for automatic deduction of the parking fee. This system is an application of Optical character recognition technology in which image is pre processed to localize only the number plate region and an algorithm is applied to extract the characters. After recognition of the license number, it is linked to the vehicle owner's account to deduct the parking fee.

This paper presents a novel hybrid method for
extracting license plates and recognizing characters
from the digital camera image using morphological
operations. The main problem in extracting text from
the images is caused by several reasons including font
size variation, alignment of text and variation of font
colors. A novel hybrid method is proposed here for
character segmentation and recognition. The objective
of this work is to design and implement algorithms for
Recognition of Indian License Plates. This paper
presents a robust algorithm of license plate location,
segmentation and reorganization of the characters
present in the located plate. Since text regions in license
plate images contain mostly repetition of vertical
strokes (edges), once the group of edges is found,
neighboring edges of a segment are connected to each
other. The proposed method is effective for slanted or
curved characters compared to existing methods.
Experimental results show that the proposed method
improves the effectiveness and robustness for License
plate recognition.
Keywords— image processing, license plate
localization, segmentation, text detection.

Письменные источники - это документы, с которыми связана история как наука. Документы истории и история документов тесно переплетены. Автоматический контент-анализ и компьютерная обработка исторических текстов обрели свое собственное место в мире методов источниковедения и конкретно-исторического анализа.

Werden für eine Liegenschaftsvermessung historische Katasterunterlagen verwendet, ist es oft notwendig, die handschriftlichen Aufzeichnungen der früheren Berufskollegen zu entziffern. Was vor einigen Jahrzehnten vielen noch ohne allzu große Mühe möglich war, bereitet heute Schwierigkeiten, sodass sich die Frage nach technischer Unterstützung durch künstliche Intelligenz stellt.

The technologies underlying fire and smoke detection systems play a crucial role in ensuring and delivering optimal performance in modern surveillance environments. In fact, fire can cause significant damage to lives and properties. Considering that the majority of cities have already installed camera-monitoring systems, this encouraged us to take advantage of the availability of these systems to develop cost-effective vision detection methods. However, this is a complex vision detection task from the perspective of deformations, unusual camera angles and viewpoints, and seasonal changes. To overcome these limitations, we propose a new method based on a deep learning approach, which uses a convolutional neural network that employs dilated convolutions. We evaluated our method by training and testing it on our custom-built dataset, which consists of images of fire and smoke that we collected from the internet and labeled manually. The performance of our method was compared with that of methods based on well-known state-of-the-art architectures. Our experimental results indicate that the classification performance and complexity of our method are superior. In addition, our method is designed to be well generalized for unseen data, which offers effective generalization and reduces the number of false alarms.