Image recognition sample application
When it comes to building an open-ended application, you want to start by defining the requirements for an MVP (short for, Minimum Viable Product) version that contains just enough functionalities to make it usable and valuable to your users. When it comes to making technical decisions for your implementation, making sure that you get a working end-to-end implementation as quickly as possible, without investing too much time, is a very important criteria. The idea is that you want to start small so that you can quickly iterate and improve your application.
For the MVP of our image recognition sample application, we'll use the following requirements:
- Don't build a model from scratch; instead, reuse one of the pretrained generic convolutional neural network (CNN: https://en.wikipedia.org/wiki/Convolutional_neural_network) models that are publicly available, such as MobileNet. We can always retrain these models later with custom training...
