Deep Learning for Computer Vision

Object detection training pipeline

The training protobuf has to be configured for training. The following five things are important in this process:

• The model configuration with the type of model
• The train_config for standard training parameters 
• The eval_config for the metrics that have to be reported
• The train_input_ config for the dataset
• The eval_input_ config for the evaluation dataset

We will use the config file from https://github.com/tensorflow/models/blob/master/research/object_detection/samples/configs/ssd_mobilenet_v1_pets.config. Download it to the Chapter04  folder by running the following command. Open the config file and edit the following lines:

fine_tune_checkpoint: "~/Chapter04/ssd_mobilenet_v1_coco_11_06_2017/model.ckpt" ...