Abstract: In this talk, we will explore how action recognition can be applied to human monitoring by analyzing video streams. Existing work has either focused on simple activities in real-life scenarios or on recognizing more complex activities (in terms of visual variability) in hand-clipped videos with well-defined temporal boundaries. However, there remains a gap in methods capable of retrieving multiple instances of complex human activity in a continuous (untrimmed) video flow in real-world settings. We will first present various techniques for detecting and tracking individuals in different environments. We will discuss various modalities, such as skeleton, optical flow, and emotion recognition, that can aid in the activity recognition process. We will then review state-of-the-art models for activity recognition and detection, including those using self-attention, transformers, and different pre-training methods. We will also cover specific cases of activity detection, such as video anomaly detection using weakly-supervised methods. Then, we will discuss several new techniques for recognizing Activities of Daily Living (ADLs) using video cameras. The proposed activity monitoring approaches will be illustrated through several home-care application datasets, including Toyota SmartHome, NTU-RGB+D, Charades, and Northwestern UCLA. This comprehensive talk will provide a thorough understanding of the current state and future directions of action detection for human monitoring in various real-world contexts.