| Titre : | Analysing Sentiments in Images with Convolutional Neural Networks. | | Type de document : | projet fin études | | Auteurs : | MAIMOUN Bakr, Auteur | | Année de publication : | 2017 | | Langues : | Anglais (eng) | | Catégories : | Ingénierie de web et Informatique mobile
| | Mots-clĂ©s : | Computer Vision, Deep Learning, Computer Science, Convolutional Neural Networks (CNNs). Dropout, Regularizations, Activation Function, layer width, network depth, pooling, hyper parameters. | | Index. dĂ©cimale : | 1722/17 | | RĂ©sumĂ© : | The passion of human-being to invent intelligent systems becomes more and more meaningful, as the data captured every second by artificial sensors needs to be examined and classified for many applications. The processing of ever-increasing amount of data by defining information explicitly seems nearly impossible, regarding the variability and the amount of the information, which reveals the need for intelligent systems that are capable of learning. Deep learning is a set of algorithms that attempts to find a hierarchical representation of the input data by trying to mimic the way human brain captures the critical aspects of excessive sensory data, to which it is exposed to every second. Convolutional neural networks, which are trainable learning structures, are also biologically inspired from the receptive fields in visual cortex. That is where computer vision using Deep Learning, idea is rapidly increasing computer science field. Generally, computer vision is considered a central aid in automation of processes where humans historically have been in control but machines could do it more efficiently. The purpose of this project is to create a program capable of recognizing sentiments from human faces in images, the feature has to detect human faces, and then analyze their sentiments one by one. The reason behind, is to deliver some relevant information to big companies about how their brand image is marketed in the social media. Hence the program should also detect brands in images. In order to support the result a thorough theoretical background to CNNs in general is provided. Here is the process we will be taking, the main task is to create a program taking an image from social media as input and giving: is there a brand in this image, is there a human face or a group of human faces and if there is some, what sentiment each face is expressing. To do this a vast and diverse data set needs to be set - both for human facial expressions, and for brands -, a comparison between some CNNs architectures to pick the right approach, an efficient and accurate network structure built (layer width, network depth, activation function, pooling…), trained and calibrated (dropout, regularizations, output layer, hyper parameters…), and a user interface built to present the results.
|
Analysing Sentiments in Images with Convolutional Neural Networks. [projet fin études] / MAIMOUN Bakr, Auteur . - 2017. Langues : Anglais ( eng) | Catégories : | Ingénierie de web et Informatique mobile
| | Mots-clĂ©s : | Computer Vision, Deep Learning, Computer Science, Convolutional Neural Networks (CNNs). Dropout, Regularizations, Activation Function, layer width, network depth, pooling, hyper parameters. | | Index. dĂ©cimale : | 1722/17 | | RĂ©sumĂ© : | The passion of human-being to invent intelligent systems becomes more and more meaningful, as the data captured every second by artificial sensors needs to be examined and classified for many applications. The processing of ever-increasing amount of data by defining information explicitly seems nearly impossible, regarding the variability and the amount of the information, which reveals the need for intelligent systems that are capable of learning. Deep learning is a set of algorithms that attempts to find a hierarchical representation of the input data by trying to mimic the way human brain captures the critical aspects of excessive sensory data, to which it is exposed to every second. Convolutional neural networks, which are trainable learning structures, are also biologically inspired from the receptive fields in visual cortex. That is where computer vision using Deep Learning, idea is rapidly increasing computer science field. Generally, computer vision is considered a central aid in automation of processes where humans historically have been in control but machines could do it more efficiently. The purpose of this project is to create a program capable of recognizing sentiments from human faces in images, the feature has to detect human faces, and then analyze their sentiments one by one. The reason behind, is to deliver some relevant information to big companies about how their brand image is marketed in the social media. Hence the program should also detect brands in images. In order to support the result a thorough theoretical background to CNNs in general is provided. Here is the process we will be taking, the main task is to create a program taking an image from social media as input and giving: is there a brand in this image, is there a human face or a group of human faces and if there is some, what sentiment each face is expressing. To do this a vast and diverse data set needs to be set - both for human facial expressions, and for brands -, a comparison between some CNNs architectures to pick the right approach, an efficient and accurate network structure built (layer width, network depth, activation function, pooling…), trained and calibrated (dropout, regularizations, output layer, hyper parameters…), and a user interface built to present the results.
|
|
172/00 
Faire une suggestion Affiner la recherche
Analysing Sentiments in Images with Convolutional Neural Networks. / MAIMOUN Bakr
