Skip to content

After training a CNN on a downsized VGGNet, gradcam had been used to visualze learned patterns from our own wallpaper dataset.

License

Notifications You must be signed in to change notification settings

navid0308/CNN-knowledge-evaluation

Repository files navigation

CNN-knowledge-evaluation

Designed a Convolutional Neural Network to learn from a million wallpaper images. Test accuracy of over 90% was achieved after 36 epochs over the 17 unique classes. Class Activation Maps were used to check if the correct patterns were learned. A few patterns such as ‘CMM’ or ‘PG’ were learned perfectly.

About

After training a CNN on a downsized VGGNet, gradcam had been used to visualze learned patterns from our own wallpaper dataset.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages