Automated recognition of violent activities from videos is vital for public safety, but often raises significant privacy concerns due to the sensitive nature of the footage. Moreover, resource constraints often hinder the deployment of deep learning-based complex video classification models on edge devices. With this motivation, this study aims to investigate an effective violent activity classifier while minimizing computational complexity, attaining competitive performance, and mitigating user data privacy concerns. We present a lightweight deep learning architecture with fewer parameters for efficient violent activity recognition. We utilize a two-stream formation of 3D depthwise separable convolution coupled with a linear self-attention mechanism for effective feature extraction, incorporating federated learning to address data privacy concerns. Experimental findings demonstrate the model’s effectiveness with test accuracies from 96% to above 97% on multiple datasets by incorporating the FedProx aggregation strategy. These findings underscore the potential to develop secure, efficient, and reliable solutions for violent activity recognition in real-world scenarios.

The pair-wise Markov chain (PMC) model serves as an extension to the hidden Markov chain (HMC) model and has been widely used in unsupervised restoration tasks associated with reconstructing the hidden data. In fact, the PMC model can treat fairly complicated situations for which application of Bayesian restoration estimators such as maximum A Posteriori (MAP), or maximal Posterior mode (MPM) remains possible. The major novelty in this work is to construct a PMC model with observational data in two dimensions, and subsequently adapt the estimation algorithms, as well as, image restoration methods for that context. Often, the transformation of an image from a two-dimensional format to a one-dimensional sequence occurs via Hilbert-Peano scan (HPS), whereas in the proposed model, the second component of the observed process takes over this role to exceed the situation of pixel missing information after transformation for a to be segmented image. To reconstruct the hidden process, we used the MPM decision criterion after estimating the model's parameters with two algorithms: Stochastic expectation maximization (SEM) and iterative conditional estimation (ICE). In this study, experimental, numerical, and visual results are shown to demonstrate the superiority of the proposed model over the classical PMC for unsupervised restorations.