Skip to content

questions about optical flow of training dataset? #7

Open
@EMJian

Description

Hello, I am deeply appreciative of your great work. While utilizing your data generation module, I successfully replicated your dataset; however, I still have a few unresolved questions:

1. I have extracted the '/cam0/optic_flow' from the bag file, which is intended to represent optical flow between consecutive frames. My confusion arises from the observation that optical flow (2hw) matches grayscale images (h*w). It has only one timestamp for the optical flow data. This leaves me uncertain as to how to ascertain which pair of consecutive frames a given optical flow corresponds to.

2. Regarding the use of temporal consistency loss, my understanding is that we employ the ground truth optical flow to warp the previously reconstructed image to the current time step. Following this warping, we compute the loss between the result of the warped image and the currently reconstructed image. Could you kindly confirm if my interpretation is accurate, or if there are any misconceptions that need correcting?

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions