Clarifai won the ImageNet 2013 with the top5 results classification error rates of 0.112 with outside data and 0.117 without. The best results of the year before was 0.153 with extra data and 0.164 without.

Therefore, it is highly desirable to implement the deep convolutional neural network architecture proposed by the paper “Visualizing and Understanding Convolutional Networks”.

Related issue: #25