The paper introduces a physics-informed deep-learning framework called Hidden Fluid Mechanics (HFM) to extract velocity and pressure fields from flow visualizations. HFM encodes the Navier-Stokes equations into neural networks, allowing for the extraction of quantitative information from flow visualizations, even in complex geometries and with low-resolution or noisy data. The method is demonstrated on various physical and biomedical problems, including external flows past a cylinder and internal flows in a patient-specific intracranial aneurysm. HFM shows robustness to low resolution and noise, making it suitable for applications in engineering and biomedicine, such as quantifying hemodynamics in vascular diseases. The framework is flexible and can be extended to other fields, such as electromagnetics.The paper introduces a physics-informed deep-learning framework called Hidden Fluid Mechanics (HFM) to extract velocity and pressure fields from flow visualizations. HFM encodes the Navier-Stokes equations into neural networks, allowing for the extraction of quantitative information from flow visualizations, even in complex geometries and with low-resolution or noisy data. The method is demonstrated on various physical and biomedical problems, including external flows past a cylinder and internal flows in a patient-specific intracranial aneurysm. HFM shows robustness to low resolution and noise, making it suitable for applications in engineering and biomedicine, such as quantifying hemodynamics in vascular diseases. The framework is flexible and can be extended to other fields, such as electromagnetics.