用来训练人头识别模型的正负样本数据集，正样本数据已经resize化。(The positive and negative sample data set is used to train the head recognition model, and the positive sample data has been resize.)