Heterogeneous Face Recognition Using Domain Specific Units


The job of Heterogeneous Face Recognition is to match facial pictures sensed in distinct domains, such as drawings for pictures (visual spectrum pictures), heat pictures for pictures, or near-infrared pictures for pictures. In this job, we propose that high-level characteristics of Deep Convolutionary Neural Networks trained on visual spectrum pictures are possibly autonomous domain and can be used to encode sensed faces in various picture domains. A generic framework for heterogeneous face recognition is suggested by adjusting low-level characteristics of Deep Convolutionary Neural Networks to "Domain-Specific Units."The adaptation using Domain Specific Units allow the learning of shallow feature detectors specific for each new image domain. Furthermore, it handles its transformation to a generic face space shared between all image domains. The Domain Specific Units adaptation enables the learning of particular shallow feature detectors for each fresh picture domain. It also handles its conversion into a generic face room shared among all picture domains. 


•Synthetic pictures using multiple face recognition algorithms such as Eigenfaces, Fisherfaces, dual space LDA and Random Sampling LDA with three photo-sketch databases  (CUHK, XM2VTS and AR database).
•Identification was carried out using the embedding network of the Visual Geometry Group (VGG) and achieved an average error rate of 34.58%.


•They suggested a filter teaching strategy with the objective of finding a convolutionary filter α, where the pixel distinction between pictures from distinct modes is the minimum
•Every work using this database implements a distinct way of reporting outcomes


•Our approach includes learning from the source domain together with the DCNN for each target domain. 
•The source domain picture is carried through the primary network and the target domain image is first passed on to its domain-specific set of feature detectors and then modified to the primary network.
•For the sake of brevity, we present the Cumulative Match Characteristic plots (CMC) for the best performed system.


•The recognition rate began to decline at the same time as the amount of free parameters began to increase exponentially.
•The best configuration is the model trained with Siamese Neural Networks on the basis of the Inception Resnet v2 and the identification rate obtained


A Deep neural convolution network (DCNN) is made up of many layers of neural networks. Typically, alternate two distinct kinds of layers, convolutional and pooling. Each filter s depth rises in the network from left to right.


Module 1:  Input datasets
In this project, we are using input data’s like sketches, thermal images or infrared images.

Module 2: Pre-processing
The aim of pre-processing is an improvement of the image data that suppresses undesired distortions or enhances some image features relevant for further processing and analysis task. Image pre-processing use the redundancy in images 

Module 3: Feature extraction
Feature extraction starts from an initial set of measured data and builds derived values intended to be informative and non-redundant, facilitating the subsequent learning and generalization steps, and in some cases leading to better human interpretations. Feature extraction is related to dimensionality reduction.

Module 4: Classification
DCNN (deep convolution neural network) is used as a classifier by using the trained model. This model designed to detect structures in streams of different images, unlike feed forward dcnn that performs computationsunidirectionally from input to output



Operating System:   WINDOWS
Simulation Tool:    OPENCV PYTHON
Documentation               :Ms-Office


CPU type              :    Intel Pentium 4
Clock speed          :    3.0 GHz
Ram size               :    512 MB
Hard disk capacity :    80 GB
Monitor type           :    15 Inch colour monitor
Keyboard type       :     Internet keyboard
CD -drive type       :     52xmax


1. Cai X, Wang C, Xiao B, Chen X, Lv Z, Shi Y (2013) Coupled latent least squares regression for heterogeneous face recognition. IEEE Inter Conf Image Process:2772–2776
2. Chen M, Fridrich J, Goljan M, Luks J (2008) Determining image origin and integrity using sensor noise. ¨ IEEE Trans Inf  Forensics Secure 3:74–90 
3. Chingovska I, Anjos A, Marcel S (2013) Anti-spoofing in action joint operation with a verification system. In: IEEE International on Computer Vision and Pattern Recognition Workshops (CVPRW), pp 98– 104 
4. de Freitas Pereira T, Anjos A, de Martino JM, Marcel S (2012) LBP-TOP based countermeasure against face spoofing attacks. Computer Vision-ACCV Workshops: 121–132

Leave a Reply
Name E-Mail Phone