Various performance indicators are used to evaluate and compare the effectiveness of object recognition methods in solving a specific task. When evaluating the performance of these methods based on the analysis of real test data, the operator's marking of pixels belonging to an object can in many cases be performed fairly approximately. The paper proposes estimates of Accuracy, Precision, Recall, and F1-score indicators of the efficiency of recognition (classification) methods under the condition of approximate markup of test data (images). The paper also suggests the so-called normalized mean-square distance between a set of False Positive pixels and a set of pixels of an object as an indicator of the performance of recognition methods, which, unlike other indicators, allows us to estimate the distribution of False Positive pixels in an image relative to pixels of objects, which is important when evaluating and comparing the effectiveness of various recognition methods. The paper provides examples of calculating the values of the proposed indicators.