Unsuitability between validation accuracy of the model, and predictions results I'm getting

[OraArbell]

Hi
In the milstone project 1 (notebook 7 tempelete)- after the fine tuning I tried to predict for the test data in order to impelemt some comparisons accuracy scores that we have learned during the module.
In the process of getting the test data we didn't shuffle the images but steel the predictions I got were totaly diffrent than the true labels although the validation accuracvy was high (77 %).
what can be the reason for that? what did I do wrong?

[muhammadanas0716]

You might have done overfitting. Check with Daniel's notebooks and try again. Most probably overfitting.

Hope this helps 💡

[OraArbell]

I will check it ! Thanks allot Ora בתאריך שבת, 16 ביולי 2022, 17:05, מאת Muhammad Anas ‏< ***@***.***>:

…

You might have done overfitting. Check with Daniel's notebooks and try again. Most probably overfitting. Hope this helps 💡 — Reply to this email directly, view it on GitHub <#401 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AZJCS2MX3YJUL5XEZYKU4CDVUK6UJANCNFSM52MU66WA> . You are receiving this because you authored the thread.Message ID: <mrdbourke/tensorflow-deep-learning/repo-discussions/401/comments/3162224@ github.com>

[muhammadanas0716]

If it helped, mark my answer as final, so I can unlock my Github achievement.