search

RobotLocomotion / pytorch-dense-correspondence

Code for "Dense Object Nets: Learning Dense Visual Object Descriptors By and For Robotic Manipulation"
https://arxiv.org/pdf/1806.08756.pdf
Other
557 stars 133 forks source link

REGARDING CLASS CONSISTENT DESCRIPTORS #202

Closed ghost closed 4 years ago

ghost commented 5 years ago

Heyy, Firstly great work. As there wasnt much info about class consistent descriptors in the paper,I would like to how it's done the training procedure requried for that,ike how correspondences are labelled for training and all. Thankyou in advance.

peteflorence commented 5 years ago

Hi there, Section 5.3 describes the class consistent descriptors. The labeling is only provided automatically by multi-view geometry from static scenes of a single object at a time. The training procedure is exactly the same as training descriptors for a single object, except the dataset includes multiple instances from the class.

On Sun, Jul 21, 2019 at 12:25 PM richard2611 notifications@github.com wrote:

Heyy, Firstly great work. As there wasnt much info about class consistent descriptors in the paper,I would like to how it's done the training procedure requried for that,ike how correspondences are labelled for training and all. Thankyou in advance.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/RobotLocomotion/pytorch-dense-correspondence/issues/202?email_source=notifications&email_token=ABLBBKBJ7MFZITLCI3GPDJ3QASEY7A5CNFSM4IFS24J2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HAP2PIA, or mute the thread https://github.com/notifications/unsubscribe-auth/ABLBBKCVZAULUUBY3KQYJWTQASEY7ANCNFSM4IFS24JQ .