Zeblok's AI-Rover™ – Beyond Explainable AI

Zeblok's AI-Rover™ provides crucial data comprehension

Entropy-as-a-Service

Quantum entropy for a connected world

• Qosmos™ is an Entropy-as-a-Service (EaaS) offering that provides quantum random numbers as a simple web-based service to the network security layer, operating systems, embedded systems or at the network edge, providing a seamless upgrade from computational security posture to information provable security.

​

• The high throughput, low cost of upgradation and ease of integration with the existing set up makes this a very cost-effective and simple-to-deploy solution over the existing infrastructure.

• The Qosmos™ architecture is designed to be scalable, provides load balancing, and has active failover.

Computer Vision analytics to detect object-person configuration and crowd counting from Drones

Aerial

Augmenting Situational

Awareness system

We present a computer vision method for processing images from a drone-mounted camera to detect foreign objects in a scene or to detect specific person-object configurations of interest—computer vision method based on MaskRCNN, the state-of-the-art appearance-based object detector framework. We will further extend MaskRCNN with an attention mechanism to incorporate contextual cues in the detection process. It uses feature similarity and spatial relationships between semantically related entities. This approach performs well for detecting humans and human hands in images, and the advantages of this approach are explainable. In essence, a region is more likely to be a hand if there are other regions with similar skin tones, and the location of a hand is inferred by the presence of other semantically related body parts such as wrist and elbow. Crowd counting is a natural extension for this algorithm, and we've seen good results.

Computer Vision-based Real-Time Data Analytics Using Edge Computing Infrastructure with Application for Smart City

Augment Smart City Applications

Usually, traditional computer vision algorithms, such as Support Vector Machines (SVM) or edge detection, can be applied to tackle such kinds of tasks in computer vision. Given the development of deep learning and 5G networks, the exploration of more advanced machine learning, such as Variational Auto Encoders (VAE) and Generative Adversarial Networks (GAN) has become applicable for the same objectives.

​

This algorithm utilizes current deep learning techniques to inspect road surfaces and monitor road conditions in real-time. Previous works involved in such topics usually apply traditional computer vision algorithms, such as edge detection, color space segmentation, or support vector machine (SVM), which are designed to work in a centralized computing environment. This algorithm creates a novel deep learning structure, such as Variational Auto-encoder (VAE) and Generative Adversarial Networks (GANs) to detect road potholes and cracks as anomalies.

Deep Learning and Adversarial Learning in Credit Card Fraud

Augment FinTech Operations

Credit card fraud is a problem that can cost banks billions of dollars annually, leading to increased incentives among financial institutions for the development of fast, effective, and dynamic fraud detection systems. This research project addresses credit card fraud detection through a semi-supervised approach, in which clusters of account profiles are created and used for modeling classifiers. Accounts are profiled based on their behavioral trends and clustered into similar groups. Groups are further identified as distinct customer segments based on purchase characteristics such as amount, frequency, or distance. Random forest and XGBoost classifiers are trained on an entire sample and compared against classifiers trained at the transaction level across each cluster.

This research concludes that the overall weighted performance of classifiers trained at the cluster level does not significantly outperform classifiers trained on the full sample. However, this research finds that clustering can be used to find meaningful groups of account holders that also have varying fraud rates across each cluster. Additionally, some classifiers trained on specific clusters yield significant improvements in performance over the baseline, whereas classifiers for other clusters do not perform as well as the baseline. This research also concludes that the optimal classifier for a given cluster varies by cluster, highlighting the potential for further development of new classifiers, which may perform well on clusters that currently exhibit under performing models.