You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am trying to use the modelTransfer to transfer the model from "Face Detection, Pose Estimation and Landmark Localization in the Wild"(http://www.ics.uci.edu/~xzhu/face/) to PartsBasedDetector. However, it seems the it misses most of the faces. I am wondering:
-> is this algorithm equivalent(roughly,within in rounding error etc) to Zhu's mixtures of trees algorithm(after transformed the model)?
Any help is appreciated
The text was updated successfully, but these errors were encountered:
I am trying to use the modelTransfer to transfer the model from "Face Detection, Pose Estimation and Landmark Localization in the Wild"(http://www.ics.uci.edu/~xzhu/face/) to PartsBasedDetector. However, it seems the it misses most of the faces. I am wondering:
-> is this algorithm equivalent(roughly,within in rounding error etc) to Zhu's mixtures of trees algorithm(after transformed the model)?
Any help is appreciated
The text was updated successfully, but these errors were encountered: