Is there any plan to release model？

[ivlydp]

No description provided.

[ztzhang]

Hi, currently there's no specific plans for that, mostly because our models are trained per sequence, which takes ~1-2 hrs on a single card. If you need them for some specific reason, I can put it on the to do list.

[ivlydp]

Thanks for your answer. Is it possible to train one model for all video sequence？The need to train each video sequence may reduce the usefulness of the algorithm~

[ztzhang]

Hi, currently the scene flow MLP is trained per sequence, while the depth network should have the ability to generalize. Therefore, if you wish to train on multiple video sequences, you might want to make changes to the scene flow MLP. One possibility is to make it conditional on some input sequence features.