I have been waiting for this update for years ! This was keeping Syntheyes from getting used in big VFX productions and why 3DE had the upper hand in anamorphic/lidar shots IMO. But now Syntheyes is all grown up, I'm really happy. I'm stuck working with 3de for the moment and it's a daily suffering, but at least I can hope of a day when I can work again with my beloved Syntheyes on the most demanding of shows. 😂
Interesting, but I always had issues aligning meshes in SynthEyes. Even if it is a CAD model of a filmed product. And still this is quite some effort here. If there is a 3D track resembling a shape - and a corresponding 3D object, shouldn´t be it just a matter of a few clicks to make software realize this and align? We live in times when one single word can product photorealistic images - with just one click.
Licensed customers can ask for help, there are many ways to do alignments in SynthEyes. Regarding "AI", sure, more things will be possible in the future. AI systems are not known for accuracy, they're better at one-time tasks than doing something repeatedly the same way. (You can see that both in SynthEyes's NN trackers and in Tesla self-driving systems, for example.) Also keep in mind that OpenAI's cumulative investment is something over $11billion; their server time alone costs billions. That's not something that can be supported for a narrow task.
@@SynthEyesHQ OpenAI is beyond scope, of course ;) But it´s possibilities too. And there are even smaller companies achieving crazy results, for example in the field of motioncapturing humans. When aligning geometry it would be a very narrow task: recognizing landmarks within two 3D-objects and aligning them. maybe this is even possible without machine learning... I mean - tracking visual points automatically (which has been possible for a long time) works also without machine learning. I´m pretty sure such an automatch is possible already - and with no doubt it will be implemented into matchmoving software. I´d just love to see SynthEyes implementing it first! ;) wasn´t meant as a complaint, sorry
oh, and regarding data density: a way to ease it up would be to mask out areas of the 3d object and just use the areas left over for the autoalign (while the rest of the geo still exists - and just moves with the small areas)
I have been waiting for this update for years ! This was keeping Syntheyes from getting used in big VFX productions and why 3DE had the upper hand in anamorphic/lidar shots IMO. But now Syntheyes is all grown up, I'm really happy. I'm stuck working with 3de for the moment and it's a daily suffering, but at least I can hope of a day when I can work again with my beloved Syntheyes on the most demanding of shows. 😂
I feel your pain!
Another awesome release. Thanks Russ.
Can't wait to try!
Brilliant! Cheers Russ.
that is awesome!!!
Hi, Russ. how can i get the plug-in exporting syntheyes tracking data to lastest Blender,
Why does my syntheyes show cranking away and hang
Interesting, but I always had issues aligning meshes in SynthEyes. Even if it is a CAD model of a filmed product.
And still this is quite some effort here. If there is a 3D track resembling a shape - and a corresponding 3D object, shouldn´t be it just a matter of a few clicks to make software realize this and align? We live in times when one single word can product photorealistic images - with just one click.
Licensed customers can ask for help, there are many ways to do alignments in SynthEyes. Regarding "AI", sure, more things will be possible in the future. AI systems are not known for accuracy, they're better at one-time tasks than doing something repeatedly the same way. (You can see that both in SynthEyes's NN trackers and in Tesla self-driving systems, for example.) Also keep in mind that OpenAI's cumulative investment is something over $11billion; their server time alone costs billions. That's not something that can be supported for a narrow task.
@@SynthEyesHQ OpenAI is beyond scope, of course ;) But it´s possibilities too.
And there are even smaller companies achieving crazy results, for example in the field of motioncapturing humans.
When aligning geometry it would be a very narrow task: recognizing landmarks within two 3D-objects and aligning them. maybe this is even possible without machine learning... I mean - tracking visual points automatically (which has been possible for a long time) works also without machine learning.
I´m pretty sure such an automatch is possible already - and with no doubt it will be implemented into matchmoving software.
I´d just love to see SynthEyes implementing it first! ;)
wasn´t meant as a complaint, sorry
oh, and regarding data density: a way to ease it up would be to mask out areas of the 3d object and just use the areas left over for the autoalign (while the rest of the geo still exists - and just moves with the small areas)
Are you ignoring working flow between syntheyes and blender?
(There's a giant SynthEyes to Blender exporter.)
@@SynthEyesHQ is it an addon and how do I get it?
File > Export > Blender @@apeiron24
omg the breathing jajaj