Rotational invariance Computer Vision
noun phrase
Definition: A property of a model, feature representation, descriptor, or transform such that rotating the input does not change the output, or does not change it in a way that affects the intended result. In computer vision, rotational invariance is typically desirable when the same object or scene should be recognized regardless of its orientation [Fei, Deng 2024].
Example in context: “However, the obtained descriptors lack rotational invariance and often exhibit shortcomings in reverse revisits.” [Luo et al. 2024]
Synonym: rotation invariance
Related terms: invariant representation, equivariance (related but distinct)