How to interpret TensorBoard loss graph?

Question

TensorBoard

I'm using Tensorflow API for object detection and here is the loss plot from its default settings. Can you help me interpret differences of classification_loss, localization_loss and objectness_loss? Thanks!

Answer 1

You could calculate the location loss like this:

def location_loss( x, y, width, height, l_x, l_y, l_width, l_height, alpha = 0.001 ):
    point_loss = ( tf.square( l_x - x ) + tf.square( l_y - y ) ) * alpha
    size_loss = ( tf.square( tf.sqrt( l_width ) - tf.sqrt( width ) ) + tf.square( tf.sqrt( l_height ) - tf.sqrt( height ) ) ) * alpha
    location_loss = point_loss + size_loss
    return location_loss

These losses you mentioned are not part of the TensorBoard. There is another TensorBoard plugin you should check the source code and the definitions for these.