Are the test results accurate?

[apuomline]

It seems that the model does not divide the data set into a test set?
好像模型训练完后，测试时依然从原647张图像进行读取，再划分对应的测试集。但是我在训练时，查看了数据集划分的训练集与验证集长度： 可见，模型在训练时见过所有图像。那模型在测试前都见过图像了，测试结果准确吗？

[Jiaoyang45]

There are only training and validation sets, no testing set. (Or it can be understood as having only training and testing sets.) Only 80% of the data participates in training, not all images. During validation, if the train_test_split parameter random_state is kept consistent with the training, the remaining 20% of the images will participate in validation. These 20% of images do not participate in the learning process of model parameters.