The random point initialization is not as big of a deal as you make it out to be as the camera poses are still essential. Most methods to arrive at camera poses give you points "for free" as a side product. It's still neat but since you still need accurate camera pose data it doesn't make a big difference in getting rid of the tedious pre-processing.
1:27:10 if thats the case i can see how in not so many years or less ai will pickup the slack, do all the research, make a fractal of different mixing inspirations from every paper, writing code and benchmarks, etc like AlphaZero for cv and other ml pipelines (especially for those we have training sets and existing benchmarks) Great Channel 👍 Subscribed ❤
This isn't integratable into existing pipelines. (So it's either "ahead of its time" or "useless" to me.) Cube marching is the only useful tech so far.
Summary starts at 1:30:23
The random point initialization is not as big of a deal as you make it out to be as the camera poses are still essential. Most methods to arrive at camera poses give you points "for free" as a side product. It's still neat but since you still need accurate camera pose data it doesn't make a big difference in getting rid of the tedious pre-processing.
1:27:10 if thats the case i can see how in not so many years or less ai will pickup the slack, do all the research, make a fractal of different mixing inspirations from every paper, writing code and benchmarks, etc like AlphaZero for cv and other ml pipelines (especially for those we have training sets and existing benchmarks)
Great Channel 👍 Subscribed ❤
That ai glitch on the eyes looks funny and feels unneccesary.
Thanks for the information though.
This isn't integratable into existing pipelines. (So it's either "ahead of its time" or "useless" to me.) Cube marching is the only useful tech so far.