jiexiong2016 / GCNv2_SLAM

Real-time SLAM system with deep features
Other
782 stars 194 forks source link

Could the GCNv2 network forward with a single RGB frame? #23

Closed siyuada closed 5 years ago

siyuada commented 5 years ago

Thanks for sharing this nice work!

In the paper, the only place you mentioned depth information is in eq. 3, to find correct correspondence. I wonder could I just passed a single RGB frame to get keypoints and descriptions into network while in inference time?

Thanks!

jiexiong2016 commented 5 years ago

Hi, The GCNv2 is using a single RGB frame and original GCN is using a pair of consecutive RGB frames.

siyuada commented 5 years ago

Thanks for reply~