Department of Computer Science and Information Systems
Permanent URI for this collectionhttp://localhost:4000/handle/123456789/1928
Browse
5 results
Search Results
Item Model Compression Based Lightweight Online Signature Verification Framework(Springer, 2022-11) Gautam, AvinashAdvances in networking and digital technologies have led to the widespread usage of Online Signature Verification (OSV) frameworks in real-time settings to validate a user's identity. Because of the superior performance of Deep Learning frameworks, CNN-based models have been widely used for solving difficult computer vision tasks such as Object Detection, Object Segmentation, and so on. The biggest impediment to OSV adoption of CNN-based models is the growing size of CNN models. This prohibits OSV frameworks from being widely used in devices with minimal computational resources, such as mobile/ embedded devices. The newly popular topic of CNN model pruning aims to solve this problem by deleting filters and neurons that do not contribute to the model's learning. Optimal CNN-based OSV frameworks are obtained by removing the less important filters and neurons and fine-tuning the pruned networks. In line with this, we offer a novel light weight OSV architecture in which a detailed ablation research is performed to examine the contribution of each layer, and non-contributive layers are deleted based on the analysis. As a result, ideal low weight models with improved classification accuracies and the ability to be applied in real-time devices emerge. Our model's performance is thoroughly tested on three commonly used datasets: MCYT-100 (DB1), SVC, and SUSIG. In MCYT-100, SVC, and SUSIG datasets, the pruned model achieves a state-of-the-art EER of 7.98%, 3.65%, and 12.39% in the skilled-1 group, respectively. The efficiency of pruning-based OSV frameworks has been demonstrated in experiments.Item TSOSVNet: Teacher-Student Collaborative Knowledge Distillation for Online Signature Verification(CVF International Conference, 2023) Gautam, AvinashOnline signature verification (OSV) is a standardized personal authentication scheme with wide social acceptance in critical real-time applications include access control, m-commerce, etc. Even though the current advances in Deep learning (DL) technologies catalysed state-of-theart frameworks for challenging domains like computer vision, speech recognition, etc., the DL-based frameworks are voluminous with huge trainable parameters and are hard to deploy in real-time systems demanding faster inference. To adopt DL into OSV for improved performance, we propose an OSV framework made up of teacher-student collaborative knowledge distillation (TSKD) technique. A heavy Transformer based teacher is trained first and the teacher knowledge is distilled into a very lightweight Convolutional Neural Network (CNN) based student. A well trained teacher network results in an efficient deep representative feature learning by the student and results in a performance improvement. In a thorough set of experiments with three popular and standard datasets, ie, the MCYT-100, SUSIG, and SVC, TSOSVNet framework, with a CNN based student model requiring only 3266 trainable parameters results in an EER of 12.42% compared to the recent SOTA 13.38% by a model with 206277 parameters in skilled 01 category of MCYT-100 dataset. In comparison to cutting-edge CNN-based OSV models, the proposed TSOSVNet produced a state-of-the-art EER in the most of the test categories with an average of 90% lesser trainable parameters.Item COMPOSV++: Light Weight Online Signature Verification Framework Through Compound Feature Extraction and Few-Shot Learning(ACM Digital Library, 2022) Gautam, AvinashOnline Signature Verification (OSV) is a systematically used biometric characteristic to endorse the genuineness of a user to access real time applications like healthcare, m-payment, etc. Because OSV frameworks are used in real-time applications and it is difficult to acquire a sufficient number of signature samples from users, they must meet a critical requirement: they must be able to detect skilled and random signature presentation attacks effectively with fewer training signature samples and a faster response time. To meet these needs, we developed a depth wise separable (DWS) convolution-based OSV framework that realizes one/few shot learning in inference phase. In addition to it, we have designed a compound feature extraction technique, which extracts maximum seven features from a set of 100 features in MCYT-100, and 3 features from a set of 47 in case of {SVC, SUSIG} datasets. The framework uses only three to seven features per signature to resist the signature presentation attacks. We have extensively evaluated our framework, by performing thorough experiments with three datasets i.e. MCYT-100, SVC and SUSIG. The model results state of the art EER in all skilled categories of SVC and SUSIG datasets.Item Impact of Type of Convolution Operation on Performance of Convolutional Neural Networks for Online Signature Verification(Springer, 2022-11) Gautam, AvinashAn Online signature is a multivariate time series, a commonly used biometric source for user verification. Deep learning (DL) is increasingly becoming ubiquitous as a paradigm for solving problems that come with a wealth of data. Convolution has been its main workhorse. Recently, DL had marked its entry in online signature verification (OSV), a standard bio-metric method that has been mostly dealt with in traditional settings. However, embracing a DL solution to a problem requires certain issues to be tackled, viz. (i) type of convolution, (ii) order of convolution, and (iii) input representation. In this work, we experimentally analyse each of the issues mentioned above regarding OSV, and subsequently present a superior model that reports state-of-the-art (SOTA) performance on three widely used data-sets namely MCYT-100, SVC, and Mobisig. Specifically, the proposed model reports an equal error rate (EER) of 9.72% and 3.1% in Skilled_01 categories of MCYT-100 and SVC data-sets, with gains of around 4% and 3% over the next best performing methods, respectively. The experimental outcome confirms that the interrelationship between the type and order of convolution operation and the input signature representation plays a significant role in the performance of OSV frameworksItem COMPOSV: compound feature extraction and depthwise separable convolution-based online signature verification(Springer, 2022-02) Gautam, AvinashOnline signature verification (OSV) is a predominantly used verification framework, which is intended to authenticate the legitimacy of a test signature by learning the writer specific signing characteristics. The significant adoption of OSV in critical applications like E-Commerce, M-Payments, etc., emphasizes on a framework which addresses critical requirements: (1) The framework should be competent to classify a test signature with few training samples, as minimum as one per user and with the least number of features extracted per signature, and (2) The framework should accurately classify a test signature of an unseen user. Even though several OSV frameworks are proposed based on various advanced techniques, still there is a necessity for a holistic OSV framework which is able to accomplish the abovementioned requirement criteria. To realize the above requirements, we present a depthwise separable (DWS) convolution-based OSV framework which facilitate the classification of test signature samples from an unseen user. In addition to this, we introduce a novel dimensionality reduction-based feature extraction technique, which decrease the dimensionality of a set of features from 100 to 3 concerning to MCYT-330, MCYT-100 and 47 to 3 with regard to SVC, SUSIG datasets. To appraise the competence of our proposed COMPOSV framework, extensive experiments and ablation studies are conducted on four widely used datasets, i.e., MCYT-100, MCYT-330, SVC and SUSIG. The proposed framework, trained with signature samples of only 10% of users (seen), can classify the signatures of 90% of unseen users with higher accuracy than the frameworks trained with signature samples of all users.