BayesOD: Bayesian Inference for Fusing Epistemic and Aleatoric - PowerPoint PPT Presentation

BayesOD: Bayesian Inference for Fusing Epistemic and Aleatoric Uncertainty in Deep Object Detectors

• • •

Box Covariance Regression • Box Mean Regression Feature Extractor NMS Classification • Sample Object Detector NMS Clustering Statistics • MC- Dropout

• Sample Object Detector Clustering Statistics • Anchor Level Priors • Object Level Box Covariance Prior Regression Box Mean Regression Bayesian Feature Extractor Clustering Inference Classification • MC- Dropout

Anchor Level Priors Object Level Box Covariance Prior Regression Box Mean Regression Bayesian Feature Extractor Clustering Inference Classification MC- Dropout

• •

• 𝑓 0 0 1 0 0 0 𝑔 0 𝑏 1 0 0 0 𝑕 𝑐 𝑑 1 •

Anchor Level Priors Object Level Box Covariance Prior Regression Box Mean Regression Bayesian Feature Extractor Clustering Inference Classification MC- Dropout

•

•

Anchor Level Priors Object Level Box Covariance Prior Regression Box Mean Regression Bayesian Feature Extractor Clustering Inference Classification MC- Dropout

Anchor Level Priors Object Level Box Covariance Prior Regression Box Mean Regression Bayesian Feature Extractor Clustering Inference Classification MC- Dropout

• •

Anchor Level Priors Object Level Box Covariance Prior Regression Box Mean Regression Bayesian Feature Extractor Clustering Inference Classification MC- Dropout

• • •

• • • o o

o o o

o o o o

• o o

• •

• • •