Intra-inter-disciplinary approaches to address open challenges of indoor and outdoor scene for videos analysis and recognition.

When we look at recent real-world applications, such as autonomous vehicle driving, use of drones to monitor events like marathons, exhibitions, processions, sports games and to study person behaviour through multiple biometric information, one can easily see that the problem is too complex and there are several open challenges. It is noted from the literature that to address such complex open issues, a single domain-based approach is not feasible. For example, use of only image information or only textual information for addressing complex challenges is not sufficient to achieve successful results. Moreover, most existing methods do not consider issues of bias and fairness. Therefore, existing methods lack effectiveness in terms of generality, robustness, consistency, reliability, fairness and unbiasedness.

The above observations motivate us to introduce intra-inter-disciplinary approaches to develop stable and reliable systems for solving complex open challenges. Different disciplines involved include image processing, pattern recognition, networking, cybersecurity, forensic, psychology, sociology, history, physics and mathematics etc. The list of challenges are as follows:

·       Multimodal approaches for domain-independent agnostic text spotting to identify persons through bib/jersey number recognition.

·       Multimodal models for Person tracking in 3D video through text spotting and image information. 

·       AI-based models for student performance prediction and understanding through handwritten document analysis.

·       AI-based techniques for photographer identification from images, for crime analysis.

·       Compressed domain-based models for deep fake/forged text and image detection.

·       Biometric signals-based models for personality traits question answering.

AI-based models for evaluating business locations, hours through person and vehicle