Please attach prototxt for tests

Please attach caffemodel for tests

Oops, caffemodel are ignored by unless using -f. Now fixed.

Please look at this warning https://pullrequest.opencv.org/buildbot/builders/precommit_docs/builds/26251/steps/patch%20size%20opencv_extra/logs/stdio
What is the main difference between data_augmentation_2x1 and data_augmentation_8x6 models? Can we only add data_augmentation_2x1 model to avoid exceeding the memory limit?

I use 2x1 and 8x6 to test that DataAug works for both enlarging (2x1 => 4x3) and shrinking (8x6 => 4x3). If I understand correctly, both model should have only hundreds of bytes?

I'm sorry if I gave you the wrong idea. Please rebase your PR to 3.4 branch.