-
Notifications
You must be signed in to change notification settings - Fork 0
dhebert-dev/image-classifier
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
********************************************************************************************************************** CFIAR10 - 50,000 images meticulously categorized by the University of Toronto ********************************************************************************************************************** I am experiencing overfitting during training. The cnn will become very accurate with the data it has the answers to but only 70 percent accurate with data it validates itself on. ********************************************************************************************************************** 10/21/2019 - testing to see if raising the validation split to 0.2 will raise the val_accuracy above 0.7 val_accuracy dipped at the 15th epoch - 0.6369 val_accuracy totalled 0.6844 at the end of 50 epoch run FAILED 10/22/2019 - Testing to see if killing more neurons will improve accuracy above 0.7 val_accuracy totalled 0.67840 at the end of 50 epoch run FAILED 10/29/2019 - There is no difference in the max validation accuracy that the cnn will reach when the 512 neuron and 1024 neuron models are compared. Additionally there is no rise above 0.7 val_accuracy when creating and only creating a second Dense layer of 512 or 1024 neurons.
About
Convolutional Neural Net Image Classifier
Resources
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published