# Why batch normalization before upsampling is giving worse results?

I am training a model to generate images.

The model contains 5+5 layers:

Conv2D -> Upsample -> Conv2D -> Upsample -> Conv2D -> Upsample -> Conv2D -> Upsample -> Conv2D -> Upsample


I am modifying it as

Conv2D -> BatchNorm -> Upsample -> Conv2D -> BatchNorm -> Upsample -> Conv2D -> BatchNorm -> Upsample -> Conv2D -> BatchNorm -> Upsample -> Conv2D -> BatchNorm -> Upsample


I am applying the batch normalization layers just before upsampling as shown above and hence I am not getting the results that are at least comparable to the results by the model without any batch normalization layer.

Is my placement of the batch normalization layer wrong? If yes, then why?

• I don't know if it makes any difference, but it may be a good idea to tell us which specific model you're using.
– nbro
Nov 29 '21 at 12:14
• @nbro Okay, I will update today. Nov 30 '21 at 22:30