Whether the results of gray image and real-world denoising are obtained by using 'ensemble'?

[cs-hao]

Thank you. Great code. I use the pre-train model to test the results, and I find the results have a slight difference.
But if I use 'ensemble' to calculate PSNR and SSIM, the results are the same.
If you do this, I think you'd better explain it.

[MC-E]

A data enhancement strategy. Please refer to the test_x8 function in RNAN and https://github.com/cszn/DPIR/blob/c4474621c47c2015b521e477a74ec0b895838069/utils/utils_model.py#L160.