Incorporating occlusion-awareness into smoothness constraints

simonmeister / UnFlow

UnFlow: Unsupervised Learning of Optical Flow with a Bidirectional Census Loss

MIT License

294 stars 57 forks source link

Closed sahityamantravadi closed 6 years ago

sahityamantravadi commented 6 years ago

Can I ask why occlusion-awareness is used in the data loss and consistency penalty, but not the smoothness constraints?

Is this something that would be useful for me to change if re-training?

Thanks!

simonmeister commented 6 years ago

Smoothness should in principle also hold inside regions becoming occluded. It may make sense to disable smoothness at motion boundaries, though.