Vision-Based Complete Scene Understanding Using Faster Region-Convolutional Neural Network

• Published in: 2024 International Conference on Data Science and Network Security (ICDSNS) pp. 1 - 5
• Main Authors: M R, Tejonidhi, Kumar Sahoo, Santosh, B M, Manjula, Soujanya, Thota, Kandaneri Ramamoorthy, Saravanan
• Format: Conference Proceeding
• Language: English
• Published: IEEE 26-07-2024
• Subjects: Accuracy; Computer vision; fast R-CNN method; holistic video understanding; Human activity recognition; human activity recognition (HAR); human robot interaction; Network security; Neural networks; Robustness; Semantics; Social robots; Videos; Visualization
• DOI: 10.1109/ICDSNS62112.2024.10690903

The computer vision is enabling the computer to process the obj ect identification and face recognition using artificial intelligence and machine learning. In computer vision the Human activity recognition (HAR) systems plays a key factor to capture from static images or playing videos. It can capture basic activities like walking and running but complex activities is a tough challenge to capture its environment objects for recognizes. In this proposed paper work address the faster Region-Convolutional Neural Network (R-CNN) method for holistic video understanding task, to enhance the human activity recognition. The semantic level description of the scene is the key factor for holistic video environment. It recognizes the human activities with autonomous system in robot and provides the contextual knowledge of the action event. Then the vision module and the social robot are integrated for natural and realistic context-based human robot interaction. The social robots need to familiar with the surrounding of the environment to react correctly for different scenarios. The faster R-CNN method obtain the accuracy with 89% with HAR dataset. The importance of contextual understanding in human activity is highlighted in recognition system.