This repo shows a convolution neural network (CNN) built for the CIFAR-100 dataset to create an object recognizer. The CNN is meant to require less memory so it can easily be trained on a g2.2xlarge AWS GPU. The trained model is then used to create a Flask/D3 app that can read in images from a URL and predict what the object is. More information about this CNN and the app can be found on my website.
Here is a demonstration of the app, showing the CNN recognizes both the bridge and castle:
