Facial Detection Using Artificial Neural Networks (ANN)
Artificial intelligence on image processing has been gaining popularity over time. Image processing such as finger printing has been simple but face recognition has been challenging due to its multi-dimensional nature. Face recognition through artificial intelligence has several steps which include; detecting individual face, extracting unique features which would be used to identify an individual and recognize the face. Face detection is a complex process which uses multiple features that complements each other. Over the past several years, face recognition has gained much attention due to its applications on passports through use of Artificial Neural Networks (ANN). There are several ANN approaches that are used in facial detection and identification. These approaches include; Retinal Connected Neural Networks (RCNN) and Convolutional Neural Networks (CNN) among others. Each of the available approach is used to detect a certain feature on human face such as retina, skin and color. Al-Allaf argues that, there is need to come up with approach that provides best performance. The main challenge involves bringing all the available detection concepts for ease of detection.
Face recognition was developed using four approaches of ANN. The most popular model of ANN is Back Propagation Neural Network (BPNN) which makes use of Back propagation Training Algorithm to achieve the concept of deep machine learning. Feed Forward Backpropagation Neural Network (FFBPNN) consists of several layers which are inter-connected to each other with the first layer communicating directly with ANN. Similarly, the last layer of FFBPNN is used to provide the required output. The training of FFBPNN can be done using an algorithm known as BP. Some learning algorithms have been deduced to adjust ANN with aim of minimizing errors and increasing performance function. ANN training steps are; starting ANN models weights as well as bias units. Next steps involves initializing learning rate, initiating required momentum variables rates and threshold errors to a very negligible value. Similarly, testing steps are; applying one face block of 14 by 8 as an input layer neurons. Doing a computation of all ANN layers output by following ANN training process and check if output layer neuron matches any expected classes.
Artificial Intelligence through use of computer facial animation is being used in film and cartoon productions. Due to advancement in technology, it has become possible to use face recognition technique to extract some information through use of interpreter’s specific face. Development of facial animation has to go through the following steps; dubbing process of any animated character, pre-modelling of the animated character, creation of layers is performed through use of RGB colors and uploading of created file is done after texture and formatting is completed. At each stage of development, a restore point is available for recovery in cases of errors during animation play. Artificial intelligence have advanced capabilities such as manipulating data, existing knowledge and other new relationships in order to solve complex problems. The entire facial image processing involves the following; facial point, facial matrix, binary matrix, information processing, 3D modelling, texture application and finalization. To model exact facial characteristics, unique individual features such as retina, facial model and posture can be used to distinguish one face from the other.
Approaches Used in Face Detection and Identification
Face recognition is becoming an area of interest in the industry ranging from law enforcement to commercial market. Despite many challenges facing facial recognition implementation, some algorithms have been developed and have proved to be useful in multimedia industry. Implementation of security features through use of facial recognition system has four distinct features namely; face detection, image pre-processing, face recognition and data extraction. Neural network is considered as a classification that can be used for predicting both unknown and known set of data. In neural networks, input nodes take information as numeric expressions and given information is presented as activation values. The acquired information is then passed from one node to the other throughout the network. The artificial neural network for face detection are; PCA with artificial neural networks, deep convolution neural networks and Bilinear CNNs among others. These approaches tries to provide justification on the need to develop best performing algorithm in facial recognition system. All these approaches have been developed with aim of making face recognition work with real time data.
The basic concept of face recognition is to determine if there exist any face in a given set of digital image. The neural network helps in creating deep machine learning which is able to detect, recognize and verify human face through use of distinct features. Face detection in digital image is classified into; image based method, knowledge based method and feature-based method among others. Concept of image detection, identification and recognition is very complex but human being have been doing it through machine learning. It is believed that there are two major modifications in face detection; neural candidate testing the face region and neural networks used in scanning image inputs.
Artificial intelligence require representation and learning of constructed concepts. The major concept which is difficult to understand by general users of the system is the ability of the machine to learn such features. First, learning of real world situation in terms of representation by machine is very important. Convolutional networks are regarded as architecture which is able to learn through several stages of developing deep machine concepts. In order to achieve deep machine learning concepts, many multi-stage labeled architecture is required. To facilitate machine learning, a Predictive Sparse Decomposition (PSD) is used. To come up with effective machine training, several procedures are evaluated in order to come up with most effective one. Since machine learning involves both hardware and software, integration is achieved through modular and object oriented concepts. Each specific module is composed of three different classes which are used to compute outputs from any given set of inputs. The major challenge in machine learning and artificial intelligence has been to come up with the best machine learning features that can automatically learn from both labeled and unlabeled data.
Face detection is one of the major parts of the body which modern technology is focusing on to enhance security. Despite challenges such as expressions, aging and distractions from wearing glasses has been presenting major challenge. In face detection, there are several methods of recognition which includes; feature vector extraction which is usually done from some of body parts including; nose, eyes, chin and mouth. All collected information is converted into feature vector. The second method is primarily based on concepts such as information theory and principle analysis of other component methods. This concept does not depend on specific part of the face but entire image. Mall (2010) proposed some methods which makes use of image coding and decoding. Data extracted from image is encoded for easy comparison with existing database models. Information processing goes through the following stages for decoding purpose; face recognition, pre-processing and face-library formation. At face-library formation, image processing is split into training and testing dataset depending on the required instance. In case of existing image, testing is done in order to get face descriptor.
Challenges Faced in Facial Detection and Recognition
Facial detection is an area of interest in modern world. Some of the areas where face recognition is being used is security, biometric analysis and videos. To make face recognition be distinct from each other, it is important to focus on color intensity instead of chrominance. Since human face has different characteristics which are distributed unevenly on the face, self-organizing map is used to extract such information. Facial detection system is being developed to help human being achieve more security parameters than traditional concepts. Once facial recognition system has been developed, it should be able to do some checks and validations. Image processing through artificial intelligence makes some business more efficient as well as gain competitive advantage
Face detection is the initial stage in face recognition system. The main issues with face detection are; illumination, occlusion, pose and high cost of computation time. Some of the approaches in use are; feature-based method, Template-based methods and Component-based methods. Face detection involves the following aspects; data acquisition, image processing, extraction of unique features and face classification depending on either face or non-face. In face detection, feature extraction is done in order to provide very essential information which can be used to differentiate both facial and non-facial. The next aspect is deep machine learning which uses arithmetic data model to learn. Machine learning has several phenomenon such as; support vector machine, neural network and Adboost which gives an explanation of machine capability to capture and interpret facial images by decoding some useful information from databases. To make face detection easier, it is important to use optimization mechanisms to reduce selection time as well as get best features. As technology advances, face detection and boosting algorithms are more preferred in real time detection of objects.
Human being are prone in using facial expressions to convey some message. As human being makes use of face to communicate, there have been attempts to make computers mimic human being in similar manner. To make machine understand facial communication, Artificial Intelligence (AI) must be implemented through machine learning. Important to note is that, both verbal and non-verbal means of communication has two aspects; body movement, psychological and body movements are the basics in non-verbal communication. AI has made it possible to create an intelligent machine which can think, make choices, image and recognize specific patterns. Face detection is being used to sense and capture some useful information related to individual face expression and information synthesis. To make it possible to differentiate some human features, neural networks are used to train machines perform some basic functions. Learning is usually done through adjustment of some learning values in order to achieve required output. In face detection arrangements, gesture representation followed involves; image input, classification, detection and tracking, image pre-processing, classification of trained information and output.
Industrial need to have reliable identification through computer systems have created more interests in biometrics development. Biometrics data are believed to advanced security features compared to traditional methods such as natural passwords. Face detection should not be a one point step, it should be able to verify and recognize those features. In this case, face recognition is believed to have two phases; enrollments and verification phases. In face detection, pre-processing module is mainly used to eliminate or minimize some variations which are observable due to light illumination aspect. Variation elimination is usually done through photometric normalization such as equalization which targets to reduce brightness and distribute it equally on the image. Feature extraction is mainly used to extract information representing any specific face. Technique used in feature extraction are known as gradient feature extraction.
Artificial Intelligence and Computer Facial Animation in Film Production
Human face is considered to have some of the complex features that are not easy to identify. Through technological advancements, it has been discovered that some of human characteristics are geometrically measurable. Development of facial recognition system which can process data in real time require real time data inputs. Neural networks in artificial intelligence is believed to be the building blocks of human nervous system. Facial recognition is highly affected by aspects such as illumination, the position of the face as image is captured, exhibited individual expression and the size of the image. Face application system is mainly used to identify and verify individuals through use of captured and encoded digital images. To develop successful facial recognition, machine learning should implement artificial face recognition to detect and identify individuals. The rate of capturing, identifying and processing digital images is quite low due to lack of high performance algorithms to process required data.
To test the originality of my report, I used turntin.com software to evaluate my work. It is one of the best tools that I found useful to determine percentage of plagiarism in any academic work. My work showed I had developed it from scratch by reading and writing it in my own words. To avoid presenting someone else work, it is important to read and understand required content and present it in my own words. Turnitin report is composed of sources where the work has been drawn from and quoted work which represents exact words as they have been drawn from reading materials. Through use of this tool, it is possible to evaluate my creativity by translating readings to my understanding.
Reference
Al-Allaf, O. N. (2014). Review of face detection systems based artificial neural networks algorithms. arXiv preprint arXiv:1404.1292. https://arxiv.org/ftp/arxiv/papers/1404/1404.1292.pdf
Al-allaf, O. N., Tamimi, A. A., & Alia, M. A. (2013). Face recognition system based on different artificial neural networks models and training algorithms. International Journal of Advanced Computer Science and Applications, 4(6). https://thesai.org/Downloads/Volume4No6/Paper_6- Face_Recognition_System_Based_on_Different_Artificial_Neural_Networks_Models_an d_Training_Algorithms.pdf
Izario, D., Izario, B., Castro, D., & Iano, Y. (2018). Face recognition techniques using artificial intelligence for audio-visual animations. Set international journal of broadcast engineering, 3 & 5. https://set.org.br/ijbe/ed3/artigo11.pdf
Kasar, M. M., Bhattacharyya, D., & Kim, T. H. (2016). Face recognition using neural network: a review. International Journal of Security and Its Applications, 10(3), 81-100. https://www.sersc.org/journals/IJSIA/vol10_no3_2016/8.pdf
LeCun, Y., Kavukcuoglu, K., & Farabet, C. (2010, May). Convolutional networks and applications in vision. In ISCAS (Vol. 2010, pp. 253-256). https://yann.lecun.org/exdb/publis/pdf/lecun-iscas-10.pdf
Mall, A., & Ghosh, S. (2010). A neural network based face detection approach. Int. J. Computer Technology & Applications, 3(2), 823-829. https://ijcte.org/papers/213-H282.pdf
Nagi, J., Ahmed, S. K., & Nagi, F. (2008, March). A MATLAB based face recognition system using image processing and neural networks. In 4th International Colloquium on Signal Processing and its Applications (Vol. 2, pp. 83-8). https://people.idsia.ch/~nagi/conferences/cspa_face_recognition.pdf
Nagi, J., Ahmed, S. K., & Nagi, F. (2008, March). A MATLAB based face recognition system using image processing and neural networks. In 4th International Colloquium on Signal Processing and its Applications (Vol. 2, pp. 83-8). https://people.idsia.ch/~nagi/conferences/cspa_face_recognition.pdf
Navabifar, F., Emadi, M., Yusof, R., & Khalid, M. (2011). A short review paper on Face detection using Machine learning. In Proceedings of the International Conference on Image Processing, Computer Vision, and Pattern Recognition (IPCV) (p. 1). The Steering Committee of the World Congress in Computer Science, Computer Engineering and Applied Computing (WorldComp). https://pdfs.semanticscholar.org/114c/8ceba9ded2a7f733d0f3ba64f380b2bb7786.pdf
Ranawade, S. S. (2010). Face recognition and verification using artificial neural network. International Journal of Computer Applications, 1(14), 21-25. https://pdfs.semanticscholar.org/fabb/c7663676a1daf4509b89e624996dc5f826c0.pdf
Supardi, J., & Utami, A. S. (2014). Development of Artificial Neural Network Architecture for Face Recognition in Real Time. International Journal Machine Learning and Computing. https://www.ijmlc.org/papers/396-H0010.pdf
Order an Essay Now & Get These Features For Free:
Turnitin Report
Formatting
Title Page
Citation
Outline
