https://codeberg.org/jbruchon/jdupes

1. I did some development myself for a "Track Discovery for Djs"[1] project in this space of "dj music recognition" and I am wondering how are you able to handle mixtapes and dj mixes when there is a significant element of sound manipulation/distortion applied, like pitch/tempo + various effects? In my tests this totally confused the algorithms which were not designed to handle such cases.

2. Can you share which algorithm have you implemented for this project? I did read most of the research papers in this space and my preferred solution was to build upon https://github.com/JorenSix/Panako which I did.

In the space of "minimal microhouse techno" type of genre where there are often similar rhythm patterns or even tracks build up using same sample packs it proved to be more difficult to have reliable results than not.

I was investigating how Spotify and other market leaders can do track recognition and they do train ML models on the same track which has applied 100+ various different effects...

Curious to hear your thoughts...

[1] - https://rominimal.club

Comments URL: https://news.ycombinator.com/item?id=39685219

Points: 1

# Comments: 0

While you browse various job boards if there is no direct contact to a person or the email / phone is obfuscate, just find contact details online.

I often pay for various services that can give you contact details based on linkedin profile - give them a call and most often then not you will have interview shortly after.

Do the work to reach out to those people, establish rapport and you will be in a good place. If you are great professional they will make sure to keep you on their roster as an ACE candidate.

For personal documents I backup into mega separate folder copy of iCloud Drive directory on each device - de duplication handles everything.

KopiaUI for offsite backups via sftp

Which is - tracks perfectly mixed together. So, I’ve created a little software/music art project.

I crawl the internet in search of new underground, (ro) minimal sounds to be mixed live by a custom, audio processing setup that was created to emulate normal Dj.

The tracks are mixed live in an unattended way, chosen semi-randomly in sets (ordered by BPM). Ready to be discovered!

The software stack is ffmpeg+python with some pre-trained ML models for beat detection.

Track discovery happens via bot on the Telegram platform (https://t.me/ROminimal_club), where to get track ID, users are asked to rate the energy level (of whatever you see those ratings mean) of a currently playing track. This is saved as his library for his future reference via private msg.

1. Go to https://ROminimal.club website and listen to the radio stream (also live video with randomly mixed content and current dj actions -- https://io.rominimal.club/ 2. Start a chat with CrossFader bot https://t.me/xFadeBot 3. Rate the energy level of tracks in the main group to reveal ID. Enjoy! example: https://t.me/ROminimal_club/7426

## What do our fans say? (Blaz) - I like how you organised your channel. You have implemented very clever gamification. And the music is top notch. My admiration. (Escobi) - some dope bangers here, i start listening 5 hours ago and still listening. Big up for this great project! The bombs are planted so great! (Lisa) - Best thing that happened on the internet in the past 10 years!

Comments URL: https://news.ycombinator.com/item?id=36468836

Points: 2

# Comments: 0

https://www.udemy.com/course/certified-kubernetes-administra...