Question about testing

[spencerkrichevsky]

Thank you @sophiajw for making this code publicly available.

I was hoping you could provide some clarification about the output of running test.py. The testA/ and testB/ folders being assigned to dataroot each contain 378k paired patches. The direction of both training and testing was set as A -> B.

The results folder (~outputs/trial/) contains 1.5M objects such as:

• A_0/ {input.png, outputA_A_0.png, outputA_B_0.png}
• B_100/ {input.png, outputB_A_0.png, outputB_B_0.png}
• outputA_1000.png
• outputB_9999.png

Now that testing has been completed, I want to merge the output patches into a "fake stain" and compare it to the "real stain" both qualitatively and quantitatively.

I'm unsure which output images should be merged together and in what order. I presume DataLoader reads the test patches in order so the output patches just need to be sorted numerically, but I'd appreciate your insight on this.