Hands-On Deep Learning for IoT

Models training

As we mentioned earlier, we are using transfer learning that does not require training from scratch; retraining of the models with a new dataset will sufficiently work in many cases. We retrained two popular architectures or models of CNN, namely Incentive V3 and Mobilenet V1, on a desktop computer, which is replicating the city council’s server. In both models, it took less than an hour to retrain the models, which is an advantage of the transfer learning approach. We need to understand the list of key arguments before running the retrain.pyfile, which is in the code folder. If we type in our Terminal (in Linux or macOS) or Command Prompt (Windows) python retrain.py -h, we shall see a window like the following screenshot ...