scaelles / OSVOS-TensorFlow

One-Shot Video Object Segmentation
http://vision.ee.ethz.ch/~cvlsegmentation/osvos/
GNU General Public License v3.0
438 stars 132 forks source link

OSVOS-S #8

Closed bhack closed 7 years ago

bhack commented 7 years ago

Do you plan to release the new OSVOS-S with instance segmentation?

scaelles commented 7 years ago

Thanks a lot for your interest in our work! We will publish the code upon acceptance of the paper.

bhack commented 7 years ago

Where was submitted?

scaelles commented 7 years ago

We have submitted it to a journal and we are waiting for their reviews.

bhack commented 7 years ago

I cannot see in the paper if you have experimented the fine-tuning of the instance segmentation mask network using the GT of the 1st frame.

scaelles commented 7 years ago

In the paper we use the instance segmentation network out of the box without any fine-tuning.

bhack commented 7 years ago

Yes I've seen.. but have you experimented the fine-tuning? Degraded?

scaelles commented 7 years ago

In DAVIS 2016, instances from different classes are mixed in the same mask so it would be difficult to use the 1st frame gt mask to fine-tune a specific class of the instance segmentation network.

bhack commented 7 years ago

But Davis 2017 could not have this problem

scaelles commented 7 years ago

It could be, but we haven't done any experiments in that direction.

bhack commented 7 years ago

So, next extension :)