TencentYoutuResearch
Can it work with multiple images ?
Based on your architechture , can this model work with more than 2 images if i increase the decoder blocks ?
The OETR architecture is not suitable for input pairs of images containing more than two images. The original intention of designing OETR was to find the co-visible area between image pairs to improve the matching performance. If you want to obtain the co-visible area in a multi-view images, the simplest approach is to divide the images into pairs and feed them into the OETR network, and then post-process the results. If you want to directly complete the co-visible area calculation in a multi-view images end-to-end, you may refer to the work related to MVS.
Duyvuleo @.***> 于2023年3月31日周五 13:41写道：
Based on your architechture , can this model work with more than 2 images if i increase the decoder blocks ?
— Reply to this email directly, view it on GitHub https://github.com/TencentYoutuResearch/ImageMatching-OETR/issues/7, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADMPDQ4QDQ7THZHSVUP5CIDW6ZVBXANCNFSM6AAAAAAWOF64XA . You are receiving this because you are subscribed to this thread.Message ID: @.***>
