Can you clarify the origin of the pre-trainend flow estimator and depth estimator checkpoints? From what kind of training were they obtained and in what experiments were they used?
Also, in what training stages was the rigid flow fine-tuning performed?
Hey I think the flow estimator was only trained on flyingchairs and flyingthings. The depth estimator was only trained on KITTI, if I remember correctly. (too long ago).
Hello,
Can you clarify the origin of the pre-trainend flow estimator and depth estimator checkpoints? From what kind of training were they obtained and in what experiments were they used?
Also, in what training stages was the rigid flow fine-tuning performed?
Thanks!