Hello,
I am quite curious about the implementation of optical flow data generation. My goal is to apply this process on another dataset. In the paper, under Section 3.3 Inference and training details > Ground-truth correspondences, the process is roughly described. Is the pipeline also available as open source?
Update: Following the citations in that section, I came across a confusing situation about the citations. In R2D2 paper, EpicFlow (cited as 40) and DeepMatching (cited as 41) are cited. In EpicFlow paper, the authors refer to another paper as DeepMatching (cited as 34 in the EpicFlow paper). Specifically:
Deepmatching paper mentioned in R2D2:
DeepMatching: Hierarchical Deformable Dense Matching
Deepmatching paper mentioned in EpicFlow:
DeepFlow: Large Displacement Optical Flow with Deep Matching
I am guessing that you wanted to cite Deepflow instead. :) I am yet to make a deep dive into those papers, maybe they are about the exact same thing, because all authors are the same. Just wanted to share this interesting situation :)
Hello, I am quite curious about the implementation of optical flow data generation. My goal is to apply this process on another dataset. In the paper, under Section 3.3 Inference and training details > Ground-truth correspondences, the process is roughly described. Is the pipeline also available as open source?
Update: Following the citations in that section, I came across a confusing situation about the citations. In R2D2 paper, EpicFlow (cited as 40) and DeepMatching (cited as 41) are cited. In EpicFlow paper, the authors refer to another paper as DeepMatching (cited as 34 in the EpicFlow paper). Specifically: Deepmatching paper mentioned in R2D2: DeepMatching: Hierarchical Deformable Dense Matching Deepmatching paper mentioned in EpicFlow: DeepFlow: Large Displacement Optical Flow with Deep Matching I am guessing that you wanted to cite Deepflow instead. :) I am yet to make a deep dive into those papers, maybe they are about the exact same thing, because all authors are the same. Just wanted to share this interesting situation :)
Best regards, Kagan