Accuracy represents the proportion of correct predictions to incorrect predictions made by a model. It is a key metric commonly used in classification models where there is a single definitive correct answer. This is distinct from object detection tasks, where the evaluation can range from a "perfect" match to being "pretty close" or "completely wrong".
In practice, terms like "top-5 accuracy" are frequently employed. "Top-5 accuracy" measures the percentage of instances where the correct answer is among the model's top 5 most confident predictions.