Infant Body

Research on infant body analysis focuses on automatically extracting and interpreting postural information from video data to improve the early diagnosis of developmental disorders and neurological conditions like cerebral palsy and autism spectrum disorder. Current efforts leverage deep learning models, including various neural network architectures (e.g., ViTPose, MediaPipe), to estimate both 2D and 3D infant poses, often addressing challenges posed by limited datasets and the unique biomechanics of infants through techniques like domain adaptation and generative priors. These advancements enable more efficient and objective assessments of infant movement and symmetry, potentially leading to earlier interventions and improved clinical outcomes.