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.