This repository contains the Health&Gait dataset, the first that enables gait analysis using visual information without specific sensors, relying solely on cameras. The dataset includes multimodal features extracted from videos, and gait parameters and anthropometric measurements from each participant. This dataset is intended for use in health, sports and gait analysis research. Health&Gait consists of 1,564 videos of 398 participants walking in a controlled closed environment, where each video has associated the following information: 2D pose estimation of their joints by AlphaPose (JSON format files). Semantic segmentation by DensePose (PNG images). Optical flow by TVL1 and GMFlow (PNG images). Silhouette by YOLOV8 (JPEG images). Moreover, for each subject, the following data has been recorded: Anthropometric measurements. Gait parameters obtained from OptoGait and MuscleLAB. Gait parameters estimated from pose information.