Review on AI vision by Elvis Bragg

The ability to train models with much larger datasets than traditional methods allows us to achieve our goal more quickly- quicker turnaround time from first idea through implementation. Also, being able to use different kinds of inputs makes it so that we can create new types of algorithms without having to spend countless hours figuring out how to get input into existing frameworks (for example Tensorflow). I would like there be an easier way to visualize what each layer is doing within Keras/Theano layers. It's hard when you're not used to this type of framework to understand exactly why your model isn't working properly or if all parameters are correct. We have been using these tools since 2016, and they've made everything faster and improved accuracy quite dramatically across multiple projects. Training large scale image recognition tasks has become significantly less painful thanks to their products! Faster turn around times plus increased quality.