Skip to content

Driver recognition and Analysis System, that consists of: active learning module for reduction of annotated images required and training time, a motion blur detection module for identification and localization of blur to retake an image in case of blurry images, and an open set recognition module to reduce false positives and increase accuracy.

Notifications You must be signed in to change notification settings

AlrikF/Driver-Recognition-and-Analysis-System

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Driver-Recognition-and-Analysis-System

Created a recognition system that can utilize low quality images for continuous monintoring and authorization of drivers. This includes models such as active learning module that utilizes entropy to indicate which images to be annotated for the highest information gain by the oracle resulting in a higher accuracy using lower number of images. The blur detection module utilizes 2 approaches one which makes use of DCT transformation for localizing blur and the other utilizeas a Mask RCNN model to model the blur map. The open set recognition helps reduce the amount of false positive by utilizing atypical samples to model the unknown class during training that helps better represent the distribution of data

About

Driver recognition and Analysis System, that consists of: active learning module for reduction of annotated images required and training time, a motion blur detection module for identification and localization of blur to retake an image in case of blurry images, and an open set recognition module to reduce false positives and increase accuracy.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published