Video analysis tasks have been shifting from deducing the present situation to forecasting the future situation. Vision-based Action Recognition is used to deduce human actions in present state, based upon complete action executions, and action prediction to predict human actions in future state based upon incomplete action executions. These two tasks have become particularly widespread recently because of their emerging real-world applications, such as mopping detection, wiping detection, talking on phone, aggressive behaviour detection, etc.
Facial recognition is a category of biometric software that maps an individual's facial features mathematically and stores the data as a faceprint. The software uses deep learning algorithms to compare a live capture or digital image to the stored faceprint in order to verify an individual's identity. With advancements in technology, it is now a viable option for authentication as well as identification. Face recognition is one of the few biometric methods that possess the merits of both high accuracy and low intrusiveness. Its use cases include face ID, photograph tagging, smart advertisements by identifying face ID, authentication and security.
The goal of object detection is to notice or discover the presence of real-world objects within an image or video frame and to be able to tell an object apart from the static background. Object detection algorithms typically use extracted features and learning algorithms to recognize instances of an object category. It is commonly used in applications such as image retrieval, security, surveillance, and automated vehicle parking systems. Some use cases of object detection include face recognition, object tracking and anomaly detection.
Posture is a way in which a human holds his body so that there is less strain on muscles during movement. Posture detection is done by determining three dimensional orientations by tracking the movement and orientation of a body with respect to a custom axes. Its use cases include real-time posture detection (talking on the phone, aggressive behavior, etc.), kinect and 3D camera based body posture detection.
Pattern recognition is a programmed or learned ability to recognise patterns within data sets. It is a fast developing field, which underpins developments in fields such as computer vision, image processing, text and document analysis and neural networks. Machines are trained to recognize the required images based on a particular pattern like recognizing a person’s face based on a particular pattern. Video streams provide us access to rich temporal patterns which are captured using AI. These temporal patterns present us with various spatial patterns which are used in applications like activity recognition, event detection/classification, anomaly detection, activity based person identification etc.