Week 1 Assignment ssi improvment

I apologize if this question has been asked previously; I was unable to find any related discussions.

I’ve successfully completed the notebook and obtained a result that appears to be correct. However, upon submission, I received an SSI of 0.45. Initially, I suspected that I was utilizing the incorrect layers for my model, so I experimented with various combinations of layers, but unfortunately, the SSI did not improve.

Here’s a breakdown of the layers I used:

content_layers = [‘conv2d_88’]

Selected five style layers of interest

style_layers = [‘conv2d’,
‘conv2d_1’,
‘conv2d_2’,
‘conv2d_3’,
‘conv2d_4’]

I would greatly appreciate it if anyone could provide some insights into where I might be going astray. Thank you in advance for your help!

Hi @Mohammad_Sarraf_Kevi the layers to use is the easiest part in the assignment (in my opinion) and yours are correct.

If i need to guess where the problem can be i will beat forany othe place, if you want send-me your notebook in a private message and I will try to figure where is the problem.

1 Like

The trickiest parts in this assignment I think are the calculation of style and content loss, you should pay special attention when you fill the code for those parts!

1 Like

I just solved the issue by changing content_and_style_layers = content_layers + style_layers to output_layers = style_layers + content_layers

I have no idea why this would make a difference.

do you mind explaining it to me ?

Yes there is a difference, if content_layers is [0:4] and style_layers is [4:8] (in lists) then their summation is not the same both ways. And remember those layers are used for different purposes so when called in a range() the layers should be positioned in the required position.

1 Like