Emotion-Aware AI systems, behaviour perception and understanding leveraging large-scale facial data to include learning-based Facial Landmark detection; Attribute Error: ‘None Type’ object has no attribute ‘shape’. This chapter discusses the state-of-the-art advanced techniques to find these important facial landmark points such as corners of eyes; tip of nose and different lip contours used for emotion modelling. We conduct an extensive survey of traditional as well as modern localisation methods with the help of deep learning such as Active Shape Models, Active Appearance Models (AAM), Convolutional Neural Networks for heatmap regression and attention mechanism to enhance the efficacy in performing localisation. Likewise, the chapter presents classical feature requirements such as LBP and HOG as reverse roles of low-level visual processing while also describing recent deep embeddings for facial texture, shape and dynamics depiction. Emphasis is placed on cross-condition performance testing illumination, head pose and occlusion, ethnicity. We also discuss the real-time use-cases, and associated implementation challenges and optimization approaches for achieving this empirical deployment across edge-AI scenarios/concerns, small device mobile environments. Furthermore, a review of present benchmark datasets and benchmarks is included and how to implement trustworthy emotion-capable AI pipelines. Combining traditional approaches with the contemporary deep learning models the explanations in this chapter provide specific guidance to researchers aiming at building efficient and reliable emotion-recognizing systems.