C3.10 | DPM | Deformable Parts Model | Object Detection | Machine Learning | Computer Vision | EvODN
Автор: Cogneethi
Загружено: 2019-08-11
Просмотров: 8670
You will learn about some of the drawbacks of Dalal & Triggs detector for non-rigid bodies and how Deformable Parts Model (DPM) overcomes those.
Conceptually, you have separate filters for different parts of the object and combine the scores at the end. This way, if some part of an non-regid object is aligned in a different way, there are chances that it will still get detected.
------------------------
This is a part of the course 'Evolution of Object Detection Networks'.
See full playlist here: • Evolution Of Object Detection Networks
------------------------
References:
Object Detection with Discriminatively Trained Part Based Model
P.F. Felzenszwalb, R.B. Girshick, D. McAllester and D. Ramanan
This work was awarded the PASCAL VOC "Lifetime Achievement" Prize in 2010.
Images Credit -
http://cs.brown.edu/people/pfelzens/p...
http://wp.doc.ic.ac.uk/lnardi/wp-cont...
https://commons.wikimedia.org/wiki/Fi...
https://commons.wikimedia.org/wiki/Fi...
http://www.publicdomainfiles.com/show...
Copyright Disclaimer: Under section 107 of the Copyright Act 1976, allowance is made for “fair use” for purposes such as criticism, comment, news reporting, teaching, scholarship, education and research.
Доступные форматы для скачивания:
Скачать видео mp4
-
Информация по загрузке:
