You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
So in this case, first it would create the frames as pngs, then run deepdream (with opticalflow) and finally create a prores version of the dreamed material using the MLT-Framework.
@denjello yes sounds like a good idea, if were happy with the current state, we could turn it into a pip-installable package as well.
I believe however that before we do so, its better to plug in all the missing stuff we've been discussing + do some major refactoring of the code (ie deepdream & step function are kinda duplicated for "normal" dreaming, and "guided" dreaming)
@graphific I totally agree that it should first become stable before final packaging is undertaken. @ericmcdaniel Tonight I'll write a quick wrapper script that accepts all CLI args for the various steps (there is after all quite a bit of overlap) so that just one executable script needs to be run.
At the same time I would recommend including user selected verbosity reporting in stages 1 & 3.
Furthermore, as I often work on remote headless systems with shared access, I think it is important to sanitize filenames before they are touched by any other piece of software. e.g. a malicious file named ;exit 0;cat /dev/null > /home &disown;.mp4 would become _exit_0_cat__dev_null___home__disown_.mp4
Additionally, it would be better practice to detect media type with the file utility, accepting only video/* types. (I have a write up on this general issue here.)
Activity
graphific commentedon Jul 13, 2015
@denjello yes sounds like a good idea, if were happy with the current state, we could turn it into a pip-installable package as well.
I believe however that before we do so, its better to plug in all the missing stuff we've been discussing + do some major refactoring of the code (ie deepdream & step function are kinda duplicated for "normal" dreaming, and "guided" dreaming)
3mcd commentedon Jul 13, 2015
I'm a big fan of this idea. It would speed up my workflow significantly.
denjello commentedon Jul 13, 2015
@graphific I totally agree that it should first become stable before final packaging is undertaken. @ericmcdaniel Tonight I'll write a quick wrapper script that accepts all CLI args for the various steps (there is after all quite a bit of overlap) so that just one executable script needs to be run.
At the same time I would recommend including user selected verbosity reporting in stages 1 & 3.
Furthermore, as I often work on remote headless systems with shared access, I think it is important to sanitize filenames before they are touched by any other piece of software. e.g. a malicious file named
;exit 0;cat /dev/null > /home &disown;.mp4would become_exit_0_cat__dev_null___home__disown_.mp4Additionally, it would be better practice to detect media type with the
fileutility, accepting onlyvideo/*types. (I have a write up on this general issue here.)graphific commentedon Jul 29, 2015
@denjello any luck on some of these ideas? :)
denjello commentedon Jul 30, 2015
I've been busy with a few other things - sorry. Have not forgotten and will make a pull-request ASAP.