I want to use my own data set to fine-tune the model, but the optical flow obtained by directly using TV-L1 does not get a reasonable result. Then I use Froward warp but the output is three-channel, and the output of the model is two-channel. How can I solve this problem??
I want to know how to get the groundtruth of training data.I hope to get your reply.
I want to use my own data set to fine-tune the model, but the optical flow obtained by directly using TV-L1 does not get a reasonable result. Then I use Froward warp but the output is three-channel, and the output of the model is two-channel. How can I solve this problem?? I want to know how to get the groundtruth of training data.I hope to get your reply.