In the last two chapters, we learned how to set up Google Colab to train a custom object detector using SSD, R-CNN, and R-FCN with Inception and MobileNet as backbone pre-trained networks. Our network was used to detect burgers and French fries. In this section, we will learn how to do the same task using GCP. A detailed description of the work can also be found at https://medium.com/tensorflow/training-and-serving-a-realtime-mobile-object-detector-in-30-minutes-with-cloud-tpus-b78971cf1193.
I started with the preceding article but found many sections have to be streamlined and additional details need to be added to get it working on my Ubuntu PC. The following subsections provide the step-by-step process for training an object detector using the GCP.
