Mastering Computer Vision with TensorFlow 2.x

Training the model

After all of the preceding steps are complete, we are ready for the most important step—training our custom neural network.

Training the model is performed using the following five steps, also outlined in the code:

1. Specify the configuration file.
2. Specify the output model directory.
3. Specify where to send the STDERR file.
4. Specify the number of training steps.
5. Specify the number of validation steps:

!python /content/models/research/object_detection/model_main.py \ --pipeline_config_path={pipeline_fname} \ --model_dir={model_dir} \ --alsologtostderr \ --num_train_steps={num_steps} \ --num_eval_steps={num_eval_steps}

The explanation for this code is as follows:

• The pipeline configuration path is defined by pipeline_fname ...