Abstract: Video observation currently plays a significant role on a global scale. Advancements have led to a significant increase in the integration of manufactured knowledge, machine learning, and deep learning into devices. The application of the combinations and unique frameworks aids in distinguishing various questionable behaviors from the real-time analysis of photographs. The human way of behaving is the most capricious; additionally, it is extremely challenging to decide if it is suspicious or normal. In this work, we have characterized human exercises into two categories: normal and suspicious. Typical exercises incorporate sitting, strolling, running, hand waving, and so forth. Arrest, abuse, shoplifting, and so on are examples of suspicious exercises. We accomplish this arrangement by utilizing convolutional neural networks. First, we use the convolutional neural network to separate significant-level highlights from pictures. We consider the grouping of the convolutional network, eliminate the outcome of the last pooling layer, and create the final forecast. The CIFAR-100 dataset confirmed the recommended model's accuracy of 0.9796 percent.
Keywords: Suspicious action; profound learning; convolutional neural organization
|
DOI:
10.17148/IARJSET.2025.12710