Is it possible to remove the perturbation for amplified frame?

[shengwa]

Hi Tiam @12dmodel ,
I'm wondering if there is a way to remove/revert the perturbation for frameB and amplified?
Best,
Sheng

[12dmodel]

Unfortunately, you would need to regenerate the dataset. I won't have time to do that in the near future, but I will be happy to answer any questions you have about data generation.

…

On Sat, Nov 3, 2018, 4:17 PM Sheng Wang ***@***.*** wrote: Hi Tiam @12dmodel <https://github.com/12dmodel> , I'm wondering if there is a way to remove/revert the perturbation for frameB and amplified? Best, Sheng — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABtIV-OxP5Wt4yCui3IdunKckvoFHlWBks5urfnbgaJpZM4YM-tO> .

[shengwa]

That would be great! Thanks!
Another question is that in your training, did you use all the 100K images?

[12dmodel]

Yes, we did. It is possible that you can use less images. If I remember correctly, we had experiments where things worked on easy cases (e.g. the baby video) with just 10k training images. However, this is not what we extensively explored so it is safer if you can afford disk space/memory to handle 100k.

…

On Mon, Nov 5, 2018 at 11:50 AM Sheng Wang ***@***.***> wrote: That would be great! Thanks! Another question is that in your training, did you use all the 100K images? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABtIV86lLfJTQWPTjUjuVMPKO0r5gEPwks5usGxCgaJpZM4YM-tO> .

[shengwa]

I got it. And how many epochs did you use for this network training?

[12dmodel]

Sorry I missed this message. We trained for about 260k iteration at a batch size of 4, so that would be about 10-11 epochs.