Center-symmetric local binary pattern (CS-LBP), an important tool for feature extraction, which is widely used for object recognition. In this research work, a reversible watermarking technique has been introduced for...
详细信息
ISBN:
(纸本)9789811524493;9789811524486
Center-symmetric local binary pattern (CS-LBP), an important tool for feature extraction, which is widely used for object recognition. In this research work, a reversible watermarking technique has been introduced for image authentication making use of local and global feature vector generated by exploiting CS-LBP, discrete cosine transform (DCT) and discrete wavelet transform (DWT). In the embedding procedure, authentication code (AC) is generated by employing CS-LBP operator on the colored cover image using a shared secret key and local feature vector and global feature vector generated by exploiting DCT and DWT which has been stored in an INFO matrix as well as embedded within the host image. In the detection procedure, authentication code (AC) is extracted with the help of a shared secret key and verified its originality by comparing it with a shared INFO matrix. Here, the symmetric property of LBP provides a minimum change in the watermarked image which helps to enhance the quality and authenticate the watermarked image.
In this dissertation, several techniques used in face recognition have been investigated. Besides, a critical problem in face recognition which is illumination variation has been considered. In this dissertation, a fa...
详细信息
The proceedings contain 1408 papers. The topics discussed include: deep gait relative attribute using a signed quadratic contrastive loss;variational capsule encoder;the DeepHealth toolkit: a unified framework to boos...
ISBN:
(纸本)9781728188089
The proceedings contain 1408 papers. The topics discussed include: deep gait relative attribute using a signed quadratic contrastive loss;variational capsule encoder;the DeepHealth toolkit: a unified framework to boost biomedical applications;hierarchically aggregated residual transformation for single image super resolution;occlusion-tolerant and personalized 3D human pose estimation in RGB images;computing stable resultant-based minimal solvers by hiding a variable;semantic segmentation for pedestrian detection from motion in temporal domain;multi-label contrastive focal loss for pedestrian attribute recognition;DmifNet: 3D shape reconstruction based on dynamic multi–branch information fusion;visual localization for autonomous driving: mapping the accurate location in the city maze;Bayesian active learning for maximal information gain on model parameters;and cross-spectrum face recognition using subspace projection hashing.
Thin flexible plates are widely used in various mechanisms such as springs, soft robot bodies and impulse force generators. Previous study proposed a method for estimating the deformation of thin flexible plates by pr...
详细信息
ISBN:
(纸本)9781728172934
Thin flexible plates are widely used in various mechanisms such as springs, soft robot bodies and impulse force generators. Previous study proposed a method for estimating the deformation of thin flexible plates by printing asymmetric and symmetric electro-conductive patterns on them. However, previous method required resistance model of conductive pattern, and the method had estimation errors due to the resistance model. Therefore, in this paper, deformed shape estimation method of thin flexible plate without resistance model is proposed. In this method, constraint conditions such both edge angles and buckling distance are estimated using resistances of conductive pattern. Then, the deformation shape is estimated from constraint conditions. In this paper, this measurement flow is simulated. Simulation results showed that the better function for calculate the relationship between constraint conditions and resistances is second degree of multivariable polynomial. We find the deformed shape estimated from the resistance values of the conductive patterns are consistent with actual deformation shape in the simulation. It confirm that the deformation shape can be estimated using the proposed estimation method.
This paper examines the human expressive states dependent on facial pictures utilizing a few viable component extraction methods. It reproduces the K-Nearest Neighbor (k-NN) classifier to approve the adequacy of succe...
详细信息
This paper examines the human expressive states dependent on facial pictures utilizing a few viable component extraction methods. It reproduces the K-Nearest Neighbor (k-NN) classifier to approve the adequacy of successful capabilities separated from the Local Binary pattern (LBP) and Histograms of Oriented Gradients (HOG) for the said task. An examination of the strategies has been made dependent on the normal acknowledgment precision of the classifiers utilizing the calculation unpredictability as a compromise. The component extraction methods have been approved for their discriminative force under various preparations for testing information division proportions, Kappa Coefficient, and order time. The LBP has outperformed the HOG include extraction strategy with a normal precision of 79.6% yet remains computationally costly. On the contrary, the HOG method has furnished a lower characterization time with a normal precision of 59.3 % as uncovered from our outcomes.
Targeting proposing a web based checking strategy for infiltration status in MIG welding, discernible bend sound sign under halfway entrance, precarious infiltration, full infiltration and extreme entrance over the sp...
详细信息
In this work, we evaluate two different image clustering objectives, k-means clustering and correlation clustering, in the context of Triplet Loss induced feature space embeddings. Specifically, we train a convolution...
详细信息
In this work, we evaluate two different image clustering objectives, k-means clustering and correlation clustering, in the context of Triplet Loss induced feature space embeddings. Specifically, we train a convolutional neural network to learn discriminative features by optimizing two popular versions of the Triplet Loss in order to study their clustering properties under the assumption of noisy labels. Additionally, we propose a new, simple Triplet Loss formulation, which shows desirable properties with respect to formal clustering objectives and outperforms the existing methods. We evaluate all three Triplet loss formulations for K-means and correlation clustering on the CIFAR-10 image classification dataset.
For the problem that the recognition rate of human upper extremity movements is not high and existing models are prone to "dimensional disaster", a patternrecognition algorithm based on PCA-LSTM neural netw...
详细信息
Malware evolves rapidly over time, which makes existing solutions being ineffective in detecting newly released malware. Machine learning models that can learn to capture malicious patterns directly from the data play...
详细信息
Malware evolves rapidly over time, which makes existing solutions being ineffective in detecting newly released malware. Machine learning models that can learn to capture malicious patterns directly from the data play an increasingly important role in malware analysis. However, traditional machine learning models heavily depend on feature engineering. The extracted static features are vulnerable as hackers could create new malware with different feature values to deceive the machine learning models. In this paper, we propose an end-to-end malware detection framework consisting of convolutional neural network, autoencoder and neural decision trees. It learns the features from multiple domains for malware detection without feature engineering. In addition, since anti-virus products should have a very low false alarm rate to avoid annoying users, we propose a special loss function, which optimizes the recall for a fixed low false positive rate (e.g., less than 0.1%). Experiments show that the proposed framework has achieved a better recall than the baseline models, and the derived loss function also makes a difference.
The general single image super-resolution methods mainly extract features from the high-resolution (HR) space by the pre-upscaling step at the beginning of the network or from the low-resolution (LR) space before the ...
详细信息
暂无评论