Published Paper Details:

Crowded event entrances pose a significant challenge in maintaining public safety, with the potential for pushing incidents leading to injuries or even fatalities. Early detection of such incidents is crucial for timely intervention and prevention of harm. In this project, we propose a novel approach leveraging machine learning techniques, specifically the UNet architecture combined with the VGG16 convolutional neural network, to detect pushing behaviors at crowded event entrances. The UNet architecture, known for its effectiveness in image segmentation tasks, is employed to accurately delineate individuals within crowded scenes. By segmenting individuals, the model gains an understanding of the spatial layout and movement patterns within the scene. Concurrently, the VGG16 network, renowned for its feature extraction capabilities, is utilized to extract high-level features from the segmented images, capturing both global and local contextual information relevant to identifying pushing behaviors.To train the model, a dataset of annotated images depicting various crowded event scenarios, including instances of pushing, is utilized. The UNet-VGG16 model is trained using a combination of supervised learning techniques, allowing it to learn discriminative features indicative of pushing behaviors .Experimental results demonstrate the efficacy of the proposed approach in accurately detecting instances of pushing at crowded event entrances. The model achieves high precision and recall rates, indicating its potential for real-world deployment in crowd management systems. Early detection of pushing incidents enables security personnel to swiftly intervene and mitigate potential risks, thereby enhancing public safety at crowded events.