LCK+: A landmark-based facial emotion recognition dataset

We introduce LCK+, a landmark-based facial emotion recognition dataset derived from the Extended Cohn-Kanade (CK+) dataset. LCK+ consists of 981 annotated samples, each containing 68 two-dimensional facial landmark coordinates and corresponding labels for one of seven basic emotions: anger, sadness, happiness, surprise, contempt, disgust, and fear. By focusing on geometric facial representations, LCK+ offers a lightweight, interpretable alternative to raw pixel data, enabling efficient and robust emotion classification. This dataset is intended to support research in affective computing, facial expression analysis, and machine learning by providing a clean, structured benchmark for evaluating landmark-based emotion recognition algorithms.