Create Your Own Computer Vision Sandbox - KDnuggets

Together we will be creating a CNN sandbox that can collect images, augment data, and easily alter architectures for quick and nimble pivoting for a variety of projects. The process was written for someone with very little CNN building experience, so will abstract away higher-level details for the benefit of being able to get a macro view on some of the fundamental components. As well several non-essential steps have been added in order to visualize outputs and provide transparency to the process. Each of these steps were written to be directly transitioned to create a model entirely of your choice, containing any amount of classifications; dogs vs cats, country flags, hotdogs vs not hotdogs. However accuracy will vary greatly dependent on the amount of classes and similarity between them.