With the rising surge of online toxicity, automating the identification of abusive language becomes crucial for improving online discourse. This study proposes a deep learning system that efficiently uses multiple labels to classify harmful comments using bi-directional Long Short-Term Memory (LSTM) networks. By leveraging contextual information, the bi-LSTM model achieves state-of-the-art performance in classifying subtle forms of toxicity such as threats, insults, identity hate, and obscenity. The model achieves above 95\% accuracy on benchmark datasets with rigorous data processing, optimized neural architecture, and the utilization of FastText embeddings to handle words that are not in the vocabulary. This technique can automatically filter different levels of toxicity, promoting positive online interactions when integrated into online platforms. The proposed study outlines an end-to-end pipeline incorporating recent NLP advancements and deep contextualized language models to address contemporary challenges in AI-enabled content moderation.

The landscape of privacy protection within the domain of video analytics is in a constant state of evolution, driven by advances in computer vision and machine learning techniques. These advancements provide increasingly sophisticated video analysis capabilities, but they also raise the stakes in terms of personal privacy. In this review, we have conducted a thorough exploration into the intersection of video analytics and privacy preservation, focusing on two core techniques: face de-identification and background blurring. We rigorously analyze the latest advancements in face de-identification, including the nuanced approaches involving facial feature perturbation, deep learning methodologies, and the intriguing domain of generative adversarial networks, meticulously highlighting both their efficacy and inherent limitations. Simultaneously, we research numerous background blurring techniques, showcasing their remarkable ability to obscure contextual information while preserving the essential elements of the visual scene. This paper underscores the practical significance of these techniques through real-world applications in surveillance and the dynamic landscape of social platforms. In essence, our comprehensive review combines the multifaceted aspects of face de-identification and background blurring within the sphere of video analytics, thereby offering a nuanced and thorough understanding of this pivotal domain.

Monkeypox has recently emerged as a public health emergency with rising cases worldwide. Early clinical diagnosis is challenging due to symptom overlap with other diseases, but characteristic skin lesions provide distinguishing visual cues. This work proposes a deep convolutional neural network (CNN) tailored for automated monkeypox screening from lesion images. A dataset of over 3000 dermatological images was compiled, with data augmentation to enhance diversity. The CNN architecture comprised convolutional blocks for feature extraction and dense layers for classification. Rigorous training and cross-validation were conducted over 100 epochs to optimize model performance. On an unseen test set, the model achieved 86.87\% accuracy in classifying monkeypox lesions, with 94\% precision, 79\% recall and 86\% F1-score. These metrics were better than baseline models, indicating reliable screening potential. Though the model overlooked some atypical presentations, successes showcase utility for mass case-finding. As monkeypox monitoring intensifies, robust computer vision approaches can assist clinicians through explainable, real-time forecasts. Prospective validation across demographics and integration with clinical workflows is warranted before full-scale deployment. Overall, the study demonstrates deep learning’s promise in tackling the monkeypox outbreak through enhanced diagnosis.