Fig 2 - uploaded by Shreyank Gowda
Content may be subject to copyright.
Source publication
Fiducial points are points that are used as points of reference or measure. Determining of fiducial points can be a fundamental step to recognize a face. A few important fiducial points are the eyes, lip edges, nose, chin etc. Using the fiducial points we can either obtain an outline of the entire face or develop a relationship between the fiducial...
Context in source publication
Citations
... In [17] Eduardo et al proposed the use of a SVM mathematical formulation called C-SVC [18], for fiducial point detection. In [19] S N Gowda et al extended the idea proposed by [17] and used a dual classification scheme to increase accuracy of the detection of fiducial points. Adaboost was used along RBF-SVM for the approach. ...
... For this step, we use the method proposed in [19]. 11 fiducial points were obtained as can be seen in figure 2(a). ...
Affective computing is an area of research under increasing demand in the field of computer vision. Expression analysis, in particular, is a topic that has been undergoing research for many years. In this paper, an algorithm for expression determination and analysis is performed for the detection of seven expressions: sadness, anger, happiness, neutral, fear, disgust and surprise. First, the 68 landmarks of the face are detected and the face is realigned and warped to obtain a new image. Next, feature extraction is performed using LPQ. We then use a dimensionality reduction algorithm followed by a dual RBF-SVM and Adaboost classification algorithm to find the interest points in the features extracted. We then plot bezier curves on the regions of interest obtained. The curves are then used as the input to a CNN and this determines the facial expression. The results showed the algorithm to be extremely successful.
... In [17] Eduardo et al proposed the use of a SVM mathematical formulation called C-SVC [18], for fiducial point detection. In [19] S N Gowda et al extended the idea proposed by [17] and used a dual classification scheme to increase accuracy of the detection of fiducial points. Adaboost was used along RBF-SVM for the approach. ...
... For this step, we use the method proposed in [19]. 11 fiducial points were obtained as can be seen in figure 2(a). ...
Affective computing is an area of research under increasing demand in the field of computer vision. Expression analysis, in particular, is a topic that has been undergoing research for many years. In this paper, an algorithm for expression determination and analysis is performed for the detection of seven expressions: sadness, anger, happiness, neutral, fear, disgust and surprise. First, the 68 landmarks of the face are detected and the face is realigned and warped to obtain a new image. Next, feature extraction is performed using LPQ. We then use a dimensionality reduction algorithm followed by a dual RBF-SVM and Adaboost classification algorithm to find the interest points in the features extracted. We then plot bezier curves on the regions of interest obtained. The curves are then used as the input to a CNN and this determines the facial expression. The results showed the algorithm to be extremely successful.
... In [8] and [9] it was seen that speech or image input can be modified to resemble relevant input to the neural network but look/sound like complete gibberish to humans. In [10], real-time input captured through a cell phone camera can be subject to adversarial perturbations that result in complete misclassification. Authors in [11] proposed the use of manifold distance, while, [12] utilized principal components to recognize adversarial attacks. ...
Minute pixel changes in an image drastically change the prediction that the deep learning model makes. One of the most significant problems that could arise due to this, for instance, is autonomous driving. Many methods have been proposed to combat this with varying amounts of success. We propose a 3 step method for defending such attacks. First, we denoise the image using statistical methods. Second, we show that adopting multiple color spaces in the same model can help us to fight these adversarial attacks further as each color space detects certain features explicit to itself. Finally, the feature maps generated are enlarged and sent back as an input to obtain even smaller features. We show that the proposed model does not need to be trained to defend an particular type of attack and is inherently more robust to black-box, white-box, and grey-box adversarial attack techniques. In particular, the model is 56.12 percent more robust than compared models in case of white box attacks when the models are not subject to adversarial example training.
... In [20] S N Gowda et al extended the idea proposed by [18] and used a twofold portrayal intend to assemble precision of the distinguishing proof of fiducial core interests. Adaboost was used along RBF-SVM for the approach. ...
Expression or emotion analysis is a research topic under great work in the field of computer vision. There are many applications it helps to serve including drowsiness testing of driver, sentiment analysis etc. First we extract facial features by using feature extraction algorithm. Next we obtain regions of interest in the image by finding areas that have high probability of feature occurrence using a Gaussian model. Finally, we use a convolutional neural network to extract information about the expressions based on the image features. We see that the proposed algorithm produces great results in terms of accuracy of the algorithm. Time needed for the execution of the algorithm is also very less and hence this algorithm proves to be very efficient in both speed of execution and also the accuracy of execution is excellent.
... In case of the holistic approaches, face recognition is done by making use of a single feature vector, that represents the whole face image. Examples of holistic approaches are the fiducial points as proposed by Gowda et al [3], the linear discriminant analysis as proposed by Martinez et al [4], using LS-SVM as proposed by Gowda et al [5], the bayesian intrapersonal classifier as proposed by Moghaddam et al [6], and the classifiers trained by Neural networks as proposed by Rowley et al [7]. ...
Human beings produce thousands of facial actions and emotions in a single day. These come up while communicating with someone and at times even when alone. These expressions vary in complexity, intensity, and meaning. This paper proposes a novel method to predict what emotion is being expressed by analyzing the face. The algorithm, because of the speed of execution, could also be used for micro expression analysis. 11 fiducial points are taken on the image after a face recognition algorithm is used. 7 classes of images are formed. These classes are the main expressions: sadness, happiness, anger, fear, disgust, surprise and neutral. Training is done by studying the relationship between the fiducial points for each class of image. Using this relationship a new image is classified by making use of the k-means algorithm.
... Dhall et al. [18] suggest an adaptive makeup algorithm that is automatic in nature. It applies makeup based on the ethnicity of skin color and gender of the person in the image. ...
Colorization is the method of converting an image in grayscale to a fully color image. There are multiple methods to do the same. Old school methods used machine learning algorithms and optimization techniques to suggest possible colors to use. With advances in the field of deep learning, colorization results have improved consistently with improvements in deep learning architectures. The latest development in the field of deep learning is the emergence of generative adversarial networks (GANs) which is used to generate information and not just predict or classify. As part of this report, 2 architectures of recent papers are reproduced along with a novel architecture being suggested for general colorization. Following this, we propose the use of colorization by generating makeup suggestions automatically on a face. To do this, a dataset consisting of 1000 images has been created. When an image of a person without makeup is sent to the model, the model first converts the image to grayscale and then passes it through the suggested GAN model. The output is a generated makeup suggestion. To develop this model, we need to tweak the general colorization model to deal only with faces of people.
... In fact, current state-of-the-art approaches on the challenging Imagenet [1] dataset has been obtained by very deep networks [12][13][14][15]. Many complex computer vision tasks have also been shown to obtain great results on using deeper networks [16][17][18][19][20]. ...
Image classification is a fundamental application in computer vision. Recently, deeper networks and highly connected networks have shown state of the art performance for image classification tasks. Most datasets these days consist of a finite number of color images. These color images are taken as input in the form of RGB images and classification is done without modifying them. We explore the importance of color spaces and show that color spaces (essentially transformations of original RGB images) can significantly affect classification accuracy. Further, we show that certain classes of images are better represented in particular color spaces and for a dataset with a highly varying number of classes such as CIFAR and Imagenet, using a model that considers multiple color spaces within the same model gives excellent levels of accuracy. Also, we show that such a model, where the input is preprocessed into multiple color spaces simultaneously, needs far fewer parameters to obtain high accuracy for classification. For example, our model with 1.75M parameters significantly outperforms DenseNet 100-12 that has 12M parameters and gives results comparable to Densenet-BC-190-40 that has 25.6M parameters for classification of four competitive image classification datasets namely: CIFAR-10, CIFAR-100, SVHN and Imagenet. Our model essentially takes an RGB image as input, simultaneously converts the image into 7 different color spaces and uses these as inputs to individual densenets. We use small and wide densenets to reduce computation overhead and number of hyperparameters required. We obtain significant improvement on current state of the art results on these datasets as well.
... In fact, current state-of-the-art approaches on the challenging Imagenet [1] dataset has been obtained by very deep networks [12][13][14][15]. Many complex computer vision tasks have also been shown to obtain great results on using deeper networks [16][17][18][19][20]. ...
Image classification is a fundamental application in computer vision. Recently, deeper networks and highly connected networks have shown state of the art performance for image classification tasks. Most datasets these days consist of a finite number of color images. These color images are taken as input in the form of RGB images and classification is done without modifying them. We explore the importance of color spaces and show that color spaces (essentially transformations of original RGB images) can significantly affect classification accuracy. Further, we show that certain classes of images are better represented in particular color spaces and for a dataset with a highly varying number of classes such as CIFAR and Imagenet, using a model that considers multiple color spaces within the same model gives excellent levels of accuracy. Also, we show that such a model, where the input is preprocessed into multiple color spaces simultaneously, needs far fewer parameters to obtain high accuracy for classification. For example, our model with 1.75M parameters significantly outperforms DenseNet 100-12 that has 12M parameters and gives results comparable to Densenet-BC-190-40 that has 25.6M parameters for classification of four competitive image classification datasets namely: CIFAR-10, CIFAR-100, SVHN and Imagenet. Our model essentially takes an RGB image as input, simultaneously converts the image into 7 different color spaces and uses these as inputs to individual densenets. We use small and wide densenets to reduce computation overhead and number of hyperparameters required. We obtain significant improvement on current state of the art results on these datasets as well.
