How to display number of epochs in tensorflow object detection api with Faster Rcnn?

Question

I am using Tensorflow Object detection api. What I understood reading the faster_rcnn_inception_v2_pets.config file is that num_steps mean the total number of steps and not the epochs. But then what is the point of specifying batch_size?? Lets say I have 500 images in my training data and I set batch size = 5 and num_steps = 20k. Does that mean number of epochs are equal to 200 ??

When I run model_main.py it shows only the global_steps loss. So if these global steps are not the epochs then how should I change the code to display train loss and val loss after each step and also after each epoch.

Answer 1

So you are right with your assumption, that you have 200 epochs.

I had a similar problem with the not showing of loss. my solution was to go to the model_main.py file and then insert

tf.logging.set_verbosity(tf.logging.INFO)

after the import stuff. then it shows you the loss after each 100 steps. you could change the set_verbosity if you want to have it after every epoch ;)