Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

TensorFlow Object Detection API - what do the losses mean in the object detection api?

What do each for the following losses mean? (in the TensorFlow Object detection API, while training FasterRCNN based models)

Loss/BoxClassifierLoss/classification_loss/mul_1

Loss/BoxClassifierLoss/localization_loss/mul_1

Loss/RPNLoss/localization_loss/mul_1

Loss/RPNLoss/objectness_loss/mul_1

clone_loss_1

like image 395
이양규 Avatar asked Jan 05 '18 10:01

이양규


People also ask

What is loss in object detection?

The loss functions of object detection can be categorized as two sorts: the classification loss and the localization loss. The former is applied to train the classify head for determining the type of target object, and the latter is used to train another head for regressing a rectangular box to locate target object.

What is loss localization loss?

Loss Function The localization loss is a smooth L1 loss between the predicted bounding box correction and the true values. The coordinate correction transformation is same as what R-CNN does in bounding box regression.

What is Objectness loss?

Objectness Loss — Knowing One's Worth. With each box prediction is associated a prediction called 'objectness'. It comes in the place where in previous detectors like RCNN came the confidence that a region proposal contains an object, because it is multiplied by the lass score to give absolute class confidence.

What is TensorFlow object detection API?

TensorFlow object detection is a computer vision technique that detects, locates, and traces an object from a still image or video. The method allows us to recognize how the models work and provides a fuller understanding of the image or video by detecting objects.


1 Answers

The losses for the Region Proposal Network:

Loss/RPNLoss/localization_loss/mul_1: Localization Loss or the Loss of the Bounding Box regressor for the RPN

Loss/RPNLoss/objectness_loss/mul_1: Loss of the Classifier that classifies if a bounding box is an object of interest or background

The losses for the Final Classifier:

Loss/BoxClassifierLoss/classification_loss/mul_1: Loss for the classification of detected objects into various classes: Cat, Dog, Airplane etc

Loss/BoxClassifierLoss/localization_loss/mul_1: Localization Loss or the Loss of the Bounding Box regressor

like image 162
Rohit Gupta Avatar answered Sep 18 '22 09:09

Rohit Gupta