Transformation invariance is a property of a system or model where its output remains unchanged under certain transformations of the input, such as rotation, translation, or scaling. This concept is crucial in fields like computer vision and machine learning, where models need to recognize objects regardless of their orientation or position.