C2W2 - Assignment not reaching 80% accuracy

Course Q&A TensorFlow Developer Professional Convolutional Neural Networks in TensorFlow
1

Hi,

course : DeepLearning.AI TensorFlow Developer Professional Certificate

Course 2 - CNN with image augmentation

In C2W2 , Cats vs Dogs classification using Data Augmentation , should get 80% as accuracy with 15 epochs but I am getting only 78.

I have applied 90 as batch for training and 10 as batch for validation

Applied same parameter values what is shown during training videos for image augemtation.

Any help will be appreciated

Epoch 15/15

250/250 [==============================] - 165s 661ms/step - loss: 0.4544 - accuracy: 0.7835 - val_loss: 0.4166 - val_accuracy: 0.8040

2

Please start with a batch size of 32 & explore from there. Did you rescale the images as part of using ImageDataGenerator ?

3

Thanks Balaji. Finally I fixed it.

  1. Reduced number of filter layers - instead of 32,64,128… set it to 16,32,64
  2. Changed the optimizer to Adam
  3. Batch size set to 90 and 10 for training and validation
  4. Yes rescaled both training and validation images to 1./255.

Epoch 15/15
250/250 [==============================] - 166s 664ms/step - loss: 0.3843 - accuracy: 0.8279 - val_loss: 0.3312 - val_accuracy: 0.8460

3 Likes
4

Hello i have same architecture with you, how much ETA you got? do you use colab pro or not?
Thank you