Add your Insight
15 January 2021
Recent Trends in Deep Learning for Computer Vision and Visual Analytics
Computer Vision (CV) sequentially integrates Artificial Intelligence (AI) techniques to understand and evaluate the real-world data effectively in many practical applications. Here, CV surpasses human visual abilities in many areas, such as recognizing faces to process the live-action of cricket, increase the speed and accuracy of therapy response assessments, automotive self-driving vehicles, and so on. Also, CV plays an essential role in forensics, surveillance, and health care systems. In general, CV works in four necessary steps: the acquisition of real-world data, processing the data, analyzing the matching features, and identifying/classifying the target data. Besides, many smart features like augmented reality, facial, iris, and gesture recognition have taken CV already to the next level. CV can play a prominent role during any emergency or pandemic situation, in recognizing the temperatures through thermal scanners, measuring & identifying the social distance in public places, and detecting faces with masks. Over the last few years, these applications have created a massive amount of streaming data daily in various domains. Considering the existence of vast data sets, cognitive scientists use computational algorithms and neural network models to extract meaningful information from imprecise or intricate data patterns through which it performs more accurate processing tasks. As a consequence, several open-source programming libraries and standardized toolkits have become pervasive for building, training, and evaluating neural network models. In this perspective, data visualization and Visual Analytics (VA) facilitate knowledge communication and analytical reasoning to support interpretation, model explanation, debugging, generalization, make further improvement in performing classification, prediction, clustering, and association. Further, the convergence of interactive/scientific/software visualization solution with graphical representation tools provides rapid, high-quality decision-making to solve practical, real-world CV problems.
With the positive growth of automated systems, the progression of Deep Learning (DL) is getting people’s consideration in numerous fields such as information and computer science, electrical and electronics, robotics, medical industry, manufacturing domains, etc. Subsequently, DL has taken CV and VA to multiple dimensions due to its computational power and visual analytical representation. From the computing perspectives, the CV/VA possesses adaptive learning, self-organization, real-time operation, fault tolerance, and prognosis in analyzing the working mechanisms of models, finding real-time results, and visualizes the data with advanced DL algorithms and software tools. DL supports the real-world systems to achieve optimum performance by regularizing the hyper-parameters of hidden layers with suitable training and testing samples. In this context, researchers use node-link diagrams, plots/charts, and instance-based analysis in visualizing the model parameters, in high-dimensional space (3D/4D) for improving the model performances during and after the training process. Further, automatic feature learning, regularization, normalization, hierarchical learning process, and data augmentation techniques play an essential role in boosting model performance. Also, the integration of CV and VA provides an outstanding, cutting-edge technology for engineers to share and examine innovations in the digitalized environment. The purpose of the special issue helps researchers and practitioners to publish and discuss the most recent innovative ideas, trends, along with the practical challenges encountered and solutions adopted in the domain of DL for CV and VA applications.
The topics of interest include, but not limited to, the following:
- Novel CV-based frameworks for a biometric recognition system
- VA for health care systems during a pandemic situation
- Human behavior analysis using CV and VA
- VA for forensics and security-enabled applications
- CV and VA for intelligent surveillance system
- Visualizing the interpretability of DL models
- and VA for human pose estimation
- CV for interactive graphics
- Predictive VA in 3D CV
- Emerging trends in autonomous vehicle navigation and monitoring systems
- Unsupervised algorithms for learning hierarchical visual representations
- Transfer learning and statistical learning methods using CV & VA
- Real-world applications, frameworks, tools, techniques using CV&VA
- Establish and analyze real-world benchmark datasets in CV & VA
Looking to Publish your Research?
We aim to make publishing with Taylor & Francis a rewarding experience for all our authors. Please visit our Author Services website for more information and guidance, and do contact us if there is anything we can help with!
View the latest tweets from tandfphysci