Algorithm 1 EM algorithm

input: Given annotators’ label and error vector and priors for each pixel output: The probability that the annotator marks the pixel correctly priors = sum(segmentations)/len(segmentations) errors = self._m_step(segmentations,np.round(priors), segmentation_region_size, segmentations_sizes) Begin for _ in range(self.n_iter):   priors = self._e_step(segmentations, errors, priors)   errors = self._m_step(segmentations, priors, segmentation_region_size, segmentations_sizes) return priors > 0.5 End Return priors, errors