Cell-Phone-detection-in-restricted-areas

The objective of this work is to detect the cell phone and/or camera used by a person in restricted areas. The paper is based on intensive image processing techniques, such as, features extraction and image classification. The dataset of images is generated with cell phone camera including positive (with cell phone) and negative (without cell phone) images. We then extract relevant features by using classical features extraction techniques including Histogram of Oriented Gradients (HOG) and Speeded up Robust Features (SURF).The extracted features are then, passed to classifier for detection. We employ Support Vector Machine (SVM), Nearest Neighbor (K-NN) and Decision tree classifier which are already trained on our dataset of training images of persons using mobile or otherwise. Finally, the detection performance in terms of error rate is compared for various combinations of feature extraction and classification techniques. Our results show that SURF with SVM classifier gives the best accuracy.

Semi-Fragile-Watermarking-by-local-linear-binary-pattern

Cheetah-detection-from-thermal-images

Information about changes in the population sizes of wild animals is extremely important for conservation and management. Wild animal populations have been estimated using statistical methods, but it is difficult to apply such methods to large areas. To address this problem, we are trying to developed several support systems for the automated detection of cheetah detection from thermal images. In this project we are going to use the cheetah dataset provided by roboflow team, then an object detection algorithm, performed to detect object of interest from thermal remote sensing images

Contrast-quality-assessment-using-deep-learning

In this work we describe a Convolutional Neural Network (CNN) to predict image contrast quality based on human visual systems (HVS) values and without a reference image. The network consists of four convolutional layer with max pooling, one fully connected layers and the output which is considered as the quality assessment metric after being normalized to a score between 0 and 1 via Sigmoid function. Within the network structure, feature learning and regression are integrated into one optimization process, which leads to a more effective model for estimating image quality based on MOS. Three different correlation also performs between the predicted values from CNN to the MOS values to find out the relation. This approach achieves state of the art performance on the CEED2016 dataset .