pascal_voc.py 文件源码

 复制代码def _py_evaluate_segmentation(self):
        """
        This function is a wrapper to calculte the metrics for given pred_segmentation results
        :param pred_segmentations: the pred segmentation result
        :return: the evaluation metrics
        """
        confusion_matrix = np.zeros((self.num_classes,self.num_classes))
        result_dir = os.path.join(self.result_path, 'results', 'VOC' + self.year, 'Segmentation')

        for i, index in enumerate(self.image_set_index):
            seg_gt_info = self.load_pascal_segmentation_annotation(index)
            seg_gt_path = seg_gt_info['seg_cls_path']
            seg_gt = np.array(PIL.Image.open(seg_gt_path)).astype('float32')
            seg_pred_path = os.path.join(result_dir, '%s.png'%(index))
            seg_pred = np.array(PIL.Image.open(seg_pred_path)).astype('float32')

            seg_gt = cv2.resize(seg_gt, (seg_pred.shape[1], seg_pred.shape[0]), interpolation=cv2.INTER_NEAREST)
            ignore_index = seg_gt != 255
            seg_gt = seg_gt[ignore_index]
            seg_pred = seg_pred[ignore_index]

            confusion_matrix += self.get_confusion_matrix(seg_gt, seg_pred, self.num_classes)

        pos = confusion_matrix.sum(1)
        res = confusion_matrix.sum(0)
        tp = np.diag(confusion_matrix)

        IU_array = (tp / np.maximum(1.0, pos + res - tp))
        mean_IU = IU_array.mean()

        return {'meanIU':mean_IU, 'IU_array':IU_array}