label_proc = prepare_label(label_batch, tf.stack(raw_output.get_shape()[1:3]), num_classes=FLAGS.num_classes, one_hot=True) # [batch_size, h, w] raw_gt = tf.reshape(label_proc, [-1,]) indices = tf.squeeze(tf.where(tf.less_equal(raw_gt, FLAGS.num_classes - 1)), 1) gt = tf.cast(tf.gather(raw_gt, indices), tf.int32) prediction = tf.gather(raw_prediction, indices)
In prepare_label .it just resize the images of SegmentationClassAug . So ,the value of raw_gt are not in [0,20] .I don't know where deal with the label images in the process . In tensorboard ,the labels_summary are black ,except the value are in [0,20] ,
label_proc = prepare_label(label_batch, tf.stack(raw_output.get_shape()[1:3]), num_classes=FLAGS.num_classes, one_hot=True) # [batch_size, h, w] raw_gt = tf.reshape(label_proc, [-1,]) indices = tf.squeeze(tf.where(tf.less_equal(raw_gt, FLAGS.num_classes - 1)), 1) gt = tf.cast(tf.gather(raw_gt, indices), tf.int32) prediction = tf.gather(raw_prediction, indices)In prepare_label .it just resize the images of SegmentationClassAug . So ,the value of raw_gt are not in [0,20] .I don't know where deal with the label images in the process . In tensorboard ,the labels_summary are black ,except the value are in [0,20] ,