The Project «Fabula»: How to find the desired video-fragment or person in a pile of video files?
- пятница, 7 августа 2020 г. в 00:12:03
geektimes
https://habr.com/en/company/neuralnet/blog/514022/
If a person is far over 20, then he has already accumulated a huge film library of his life, as well as videos from friends, relatives, and from his place of work… It is no longer possible to find someone or something specific there. Recently, I was preparing a video compilation for my daughter's anniversary – I spent a week. The media is all the more overloaded with video archives. And every day, millions of terabytes of video content appear in the world. And this is in the era of BIG DATA.
So far, even YouTube does not allow you to make a selection for all videos, for example, where there are small dogs or your favorite actress meets. We did not wait for Google to be able to search not only for the names of the videos…
And here is our neural network for video archives — Fabula.
This is just the first sample of the pen that our novice programmers do, we always set them such unexpected tasks, because professionals are all busy making money. Therefore, neural networks are not the most super, but the main thing here is the principle itself, which allows you to turn any movie library into a search environment.
This site shows how videos are laid out:
The proposed program is downloaded by the user and scans the entire contents of the disk, creating index files for search queries. We are looking forward to any ideas, as well as programmers who will continue this and other similar projects. Say something about how to improve the project, and get a prize – free software from our company, which costs hundreds of thousands, and it has a lot of neural networks.
- Fabula corporate blog
- IPTV
- Big Data
- Display advertising
- Video equipment
If a person is far over 20, then he has already accumulated a huge film library of his life, as well as videos from friends, relatives, and from his place of work… It is no longer possible to find someone or something specific there. Recently, I was preparing a video compilation for my daughter's anniversary – I spent a week. The media is all the more overloaded with video archives. And every day, millions of terabytes of video content appear in the world. And this is in the era of BIG DATA.
So far, even YouTube does not allow you to make a selection for all videos, for example, where there are small dogs or your favorite actress meets. We did not wait for Google to be able to search not only for the names of the videos…
And here is our neural network for video archives — Fabula.
This is just the first sample of the pen that our novice programmers do, we always set them such unexpected tasks, because professionals are all busy making money. Therefore, neural networks are not the most super, but the main thing here is the principle itself, which allows you to turn any movie library into a search environment.
This site shows how videos are laid out:
- click on stories to quickly view all the video content and go to the right moment,
- on the faces of participants to quickly view all the stories with their participation,
- search items (people, cars, faces, books, phones, bottles… more than 300 objects).
The proposed program is downloaded by the user and scans the entire contents of the disk, creating index files for search queries. We are looking forward to any ideas, as well as programmers who will continue this and other similar projects. Say something about how to improve the project, and get a prize – free software from our company, which costs hundreds of thousands, and it has a lot of neural networks.