An Automated Content Base Facial Image Retrieval using Conventional Neural Networks

Project Title

An Automated Content Base Facial Image Retrieval using Conventional Neural Networks

Project Area of Specialization Artificial IntelligenceProject Summary

With the development of multimedia technology, the rapid increasing usage of wild image database becomes possible. To carry out its management and retrieval, Content-Based Image Retrieval (CBIR) is an effective method. This project shows the advantage of content-based image retrieval system, as well as key technologies. Compare to the shortcoming that only certain one feature is used in the traditional system, this project introduces a method that combines color, texture and shape for image retrieval and shows its advantage. Then this project focuses on the feature extraction and representation using Deep Neural Networks and Machine Learning Algorithms, several commonly used algorithms and image matching methods.

Project Objectives

In the recent past the advancement in computer and multimedia technologies has led to the production of digital images and cheap large image repositories. The size of image collections has increased rapidly due to this, including digital libraries, medical images etc. To tackle this rapid growth it is required to develop image retrieval systems which operates on a large scale. The primary aim is to build a robust system that creates,manages and query image databases in an accurate manner. CBFIR is the procedure of automatically indexing images by the extraction of their low-level visual features, like shape, color, and texture, and these indexed features are solely responsible for the retrieval of images. Thus, it can be said that through navigation, browsing, query-by-example etc. we can calculate the similarity between the low-level image contents which can be used for the retrieval of relevant images. Images are a representation of points in a high dimensional featurespaceandametricisusedtomeasurethesimilarityordissimilaritybetweenimages on this space. Therefore, those images which are closer to the query image are similar to it and are retrieved. Featurerepresentationandsimilaritymeasurementareverycrucialfortheretrievalperformance of a CBFIR system and for decades researchers have studied them extensively. A variety of techniques have been proposed but even then it remains as one of the most challenging problems in the ongoing CBFIR research, and the main reason for it is the semantic gap issue that exists between the low-level image pixels captured by machines and high level semantic concepts perceived by humans. Such a problem poses fundamental challenge of Arti?cial Intelligence from a high-level perspective that is how to build and train intelligent machines like human to tackle real-world tasks. One promising technique is Machine Learning that attempts to address this challenge in the long-term. In the recent years there have been important advancements in machine learning techniques. Deep Learning is an important break through technique,which includes a family of machine learning algorithms that attempt to model high-level abstractions in data by employing deep architectures composed of multiple non-linear transformations. Deep learning impersonates the human brain that is organized in a deep architecture and processes information through multiple stages of transformation and representation, unlike conventional machine learning methods that are often using shallow architectures. By exploring deep architectures to learn features at multiple level of abstracts from data automatically, deep learning methods allow a system to learn complex functions that directly map raw sensory input data to the output, without relying on human-crafted features using domain knowledge.

Project Implementation Method

Face Detection 
The aim of face detection is localization of the face in a image. In the case of video input, it can be an advantage to track the face in between multiple frames, to reduce computational time and preserve the identity of a face (person) between frames. Methods used for face detection includes: shape templates, neural networks. 
Preprocessing
The aim of the face preprocessing step is to normalize the coarse face detection, so that a robust feature extraction can be achieved. Depending of the application, face preprocessing includes: alignment (translation, rotation, scaling) and light normalization/correlation. 
Feature Extraction 
The aim of feature extraction is to extract a compact set of interpersonal discriminating geometrical or/and photometrical features of the face. Methods for feature extraction include: PCA. 
Feature Matching 
Feature matching is the actual recognition process. The feature vector obtained from the feature extraction is matched to classes (persons) of facial images already enrolled in a database. The matching algorithms vary from the fairly obvious nearest neighbor to advanced schemes like neural networks. 
 

Benefits of the Project

A content-based facial image retrieval (CBFIR) system works on the low-level visual features of a user input query image, which makes it dif?cult for the users to formulate the query and also does not give satisfactory retrieval results. In the past image annotation was proposed as the best possible system for CBFIR which works on the principle of automatically assigning keywords to images that help image retrieval users to query images based on these keywords. where images are represented by some low-level features and the mapping between low-level features and high-level concepts (class labels) is done by supervised learning algorithms. In a CBFIR system learning of effective feature representations and similarity measures is very important for the retrieval performance. Semantic gap has been the key challenge for this problem. A semantic gap exists between low-level image pixels captured by machines and the high-level semantics perceived by humans. The recent successes of deep learning techniques especially Convolutional Neural Networks (CNN) in solving computer vision applications has inspired me to work on this thesis so as to solve the problem of CBIR using a dataset of annotated images.

Technical Details of Final Deliverable

Firstly, there is a data set containing 44000+ facial images, as we have content base data so we have unique name (ID) of every person so then the labels of same person images are saved in a file with image label and specific key.

Secondly, the existing data set divided into two sections in training and testing. The sections are divided applying the check of having four or more than four same images, then one of the image save in testing section and rest of the images goes in training sections, one thing is to be noted that less than four same images will go in training section to check the accuracy of the system.

Now apply the pre-trained facial recognition model in training section images and then save each image’s feature vector, label and address in variable.

Similarly apply the pre trained facial recognition model in testing section image and get the image’s feature vector and then compare these feature vector Using Euclidean, cosine and SVM algorithm with the training images’ feature vector. Same feature images will be retrieved.

Final Deliverable of the Project Software SystemType of Industry Education , Agriculture , Others , Security Technologies Artificial Intelligence(AI), Big DataSustainable Development Goals Decent Work and Economic Growth, Partnerships to achieve the GoalRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	67900
Camera Canon 1500D DSLR	Equipment	1	52900	52900
Graphic Card (GTX 1060) 6 GB version)	Equipment	1	15000	15000