Built a custom image classifier that will train on the fly in the browser using TensorFlow.js, used transfer learning to create a highly accurate model with minimal training data. A pre-trained model for image classification called MobileNet was used. Trained a model on top of this one to customize the image classes it recognizes.