PLEASE HELP ME to : Tackle Overfitting with Data Augmentation
I am trying to solve the solution under this assignment for more than 10 days now. The goal is that my model should achieve a training and validation accuracy of at least 80% and the final testing accuracy should be either higher than the training one or have a 5% difference at maximum. I have tuned all the parameters involved, right from the image augmentation parameters, number of CNN layers and learning but still does not work. Please I need assistance and hints. Thanks

Please move your topic to the correct subcategory.
Here’s the community user guide to get started.

Hi, Ahmad.

Yes, I have the same question as Balaji: you filed this under DLS Course 4 which is about Convolutional Models. I am a mentor for that course, so I’m very familiar with the material there, and I don’t remember any assignments with the title that you list in your question. Data Augmentation does get demonstrated and used in the Transfer Learning with MobilNet assignment in DLS C4 W2, but it doesn’t sound like what you are describing.

You need to file this in the correct category to maximize your chances that a mentor for the relevant course will notice and can offer help. You can use the little “edit pencil” on the title to move it. Or if you can’t figure out how to do that, please reply with the Specialization and Course and Week for your question and I can move it for you.


Thank you for the guidance

The specialization is Convolutional Neural Networks in TensorFlow and it’s Week 2 Assignment on Cats vs Dogs with Data Augmentation.

Thank you very much Paul. Looking forward to learning more from you

Thanks for the followup info. I moved the thread to that category and I hope one of the TF mentors will be able to help.

But reading your original question again, I think it will require that you give a bit more detail in order to get meaningful help. Just saying “I’m having a problem” in effect doesn’t really give us
much to go on. For example, I were going to try to help my first question would be what the results are that you are getting with your best attempt so far. What are the training and validation accuracy values you are seeing? Do they have any test cases in the notebook that check the basic correctness of the network you are defining? In other words, do you have any way to assess whether you are making some basic mistake or is the assignment structured in a sort of “open-ended” way?

Have you seen this topic?

I sincerely appreciate your generous help and guidance, Master Paul. This is my first post in the community that was why the errors keep coming. Nevertheless, I’m delighted to have encountered such supportive mentors here. I’ll revise my question to elicit more valuable responses.

I have come across some relevant posts on the topics you posted. I will try them out. Thank you sincerely, Sir Balaji