Abstract: Methods are provided for estimating a pose of an infant using image analysis and artificial intelligence. A classifier is trained using a dataset containing hybrid synthetic and real infant pose data. Multi-stage invariant representation machine learning strategies are employed that transfer knowledge from adjacent domains of adult poses and synthetic infant images into a fine-tuned domain-adapted infant pose estimation model.

Abstract: Methods and systems for providing a dataset of human in-bed poses include simultaneously gathered images of in-bed poses of humans from imaging modalities including red-green-blue (RGB) and one or more of long wavelength infrared (LWIR), depth imaging, and pressure mapping. The images are obtained under a lighting condition and a cover condition. The dataset can be used to train a model of estimating human in-bed poses and for methods of estimating human in-bed poses. Methods and systems of estimating three-dimensional human poses from two-dimensional input images are provided.