In this article the authors introduce . . .
PixelPlayer, a system that, by watching large amounts of unlabeled videos, learns to locate image regions which produce sounds and separate the input sounds into a set of components that represents the sound from each pixel. Our approach capitalizes on the natural synchronization of the visual and audio modalities to learn models that jointly parse sounds and images, without requiring additional manual supervision.
The system is trained with a large number of videos containing people playing instruments in different combinations, including solos and duets. No supervision is provided on what instruments are present on each video, where they are located, or how they sound. During test time, the input to the system is a video showing people playing different instruments, and the mono auditory input. Our system performs audio-visual source separation and localization, splitting the input sound signal into N sound channels, each one corresponding to a different instrument category. In addition, the system can localize the sounds and assign a different audio wave to each pixel in the input video.
A video is included along with an explanation of several interesting demos, such as pointing at any pixel to hear the sound from that pixel. Or remixing the volume levels of different musical instruments in the video.
The paper is included along with the data set. It says the code is coming soon.
(Score: 4, Insightful) by fishybell on Tuesday September 25 2018, @11:53PM (2 children)
I feel like what was missing (didn't RTFA, but did WTFV) is picking out the sound of a person talking and isolate that. Even if all it did was drown out background noise, then we'd have a winner. Adding in existing filtering technologies, and we're talking not just about a potentially profitable product, but also a scary big brother tech that could isolate sounds in a crowd.
(Score: 2) by mhajicek on Wednesday September 26 2018, @12:31AM (1 child)
If I could take a YouTube video and remove the annoying music playing too loudly over the person talking that would be great. Or take a TV show or movie and cut the volume of the music and sound effects, and boost the dialogue so I can actually understand what's being said.
The spacelike surfaces of time foliations can have a cusp at the surface of discontinuity. - P. Hajicek
(Score: 1, Interesting) by Anonymous Coward on Wednesday September 26 2018, @03:33PM
Sometimes also the opposite applies: Music played in the background of some talking, and you like the music, and would like to remove the talking without affecting the quality of the music, so you can enjoy the music in isolation.