Will you publish the ground-truths of test and validation sets? It may lead to more works related to this task.