Clarification on Varying Evaluation Metrics Across Datasets in 3D Reconstruction

[wheltz]

Hello,
I'm new to the field of 3D reconstruction and have a question about why PSNR, SSIM, and LPIPS are calculated differently across various datasets. I understand that this variation often stems from the baseline methods choosing different evaluation metrics. However, when assessing the actual performance of some methods, I find many of the reported metrics in this field to be unreliable. This is because, upon reproduction by others on new datasets, it's unclear which specific evaluation methods were used. This inconsistency has been confusing for me. Is this thing in evaluation metrics a common practice in the field?