How do rcnn functions validate the trained model

I'm trying to perform rcnn detection functions (rcnn, fast-rcnn, faster-rcnn) on a custom dataset. My data has about 800 training images and 400 validation images. But the data on which the rcnn is trained is only one table and has no specification of training or validation or test sets. My question is how can I validate and test the results of training rcnn methods on a dataset like pascal VOC?

[Simon MADEC]

I guess Matlab do not use a validation dataset, the mini match accuracy is on the training dataset. But I am not sure about that