22
edits
GMU:Computer's Cut/Julian Mosbach (view source)
Revision as of 08:07, 31 January 2022
, 31 January 2022no edit summary
No edit summary |
No edit summary |
||
| Line 12: | Line 12: | ||
{{#ev:youtube|6VP3XOJsncg|740|left}} | {{#ev:youtube|6VP3XOJsncg|740|left}} | ||
I took Bachs Cello Suite No. 1 and paired it up with the Banderie from the Orchestra Suite BWV 1067. The catch is: I transposed the score so the key is a tritone above the key from the Cello suite. What that means is the impurities of the cuts lead to the most disgusting harmonics possible. But through this mess of ear-bleeding-ugliness, you can still hear the melody. Hopefully. | I took Bachs Cello Suite No. 1 and paired it up with the Banderie from the Orchestra Suite BWV 1067. The catch is: I transposed the score so the key is a tritone above the key from the Cello suite. What that means is the impurities of the cuts lead to the most disgusting harmonics possible. But through this mess of ear-bleeding-ugliness, you can still hear the melody. Hopefully. | ||
== Cats...? == | |||
{{#ev:youtube|jQhPzAkYWHQ|740|left}} | |||
As a big fan of cats, I naturally wanted to test the AIs ability to detect them. For what is the world without cats? Less hairy, probably. | |||
I knew AI is under a lot of critique lately and so I chose the easiest source material I could think of: The famous Weimar Cat Documentary, specifically the first 10 Minutes of it. What should have been the output was about a good 5 minutes of nothing but cats, but instead I sat in front of my computer watching in horror as the algorithm not only failed to identify about 95% of all the cats, but also just straight up selected a dog. And those cats weren't difficult either. I could forgive a furball in the corner of a shot not being selected, but the AI just missed frame-filling close-ups. | |||
CONCLUSION: the software was trained by dog-persons. | |||
3/10. | |||