does anyone know anything about BEM css? I have questions about how I'm supposed to do stuff but there's no slack channel for it 😆

Anyone have any good ideas on how to label frames of video for machine learning?

Each frame will have about 5 labels each of which would get its own classifier in my head right now

I just want a way to do this efficiently

I dont have to label regions or anything

just say what category the frame is in

This is a color scheme I originally put together for one of our services (Zendy), but it makes for a rather nice Slack color scheme as well: #2A313E,#3A4050,#14ABF7,#FFFFFF,#3A4050,#F7F8FB,#2DD3C4,#D91A5F

@emccue what do you mean label? Could you not just have an accompanying json file or something with the labels and frame timestamps?

@henrik that theme matches the color scheme I'm using already

@henrik https://github.com/arcticicestudio/nord-slack nord is a similar scheme

#2E3440,#3B4252,#88C0D0,#2E3440,#3B4252,#D8DEE9,#A3BE8C,#81A1C1

someone posted this clojure-ish one a few weeks ago: #293C4B,#34495E,#5CB5CD,#FFFFFF,#34495E,#FFFFFF,#7CD32B,#F2AE00

anyone can recommend a simple/cheap performance analyzer for postgresql over RDS?

I need to know which queries are being slow, or if too many reqs/sec are being performed

@jgh hypothetically that works, but i have about 10gb of video, so I cant do that manually

what are you trying to do exactly? What are you labeling?

a few hours of halo pc footage

so for a first phase in this project I am trying to label 1. What weapon am I holding

2. What is my health

3. What are my shields at

I vaguely remember talking to people who attempted to figure out the genre of a film by classifying its frames

as in, they made their living this way. maybe it was vimeo or something?

and even they didn’t analyse everything, just sampled every nth frame, and somehow modified the frequency of frame sampling by how likely they were going to find something new / relevant

@lady3janepl have you seen this? https://imgur.com/a/1q9IO

in short, I wouldn’t attempt to label every frame.

if that information is in the same place on screen constantly you could probably put together an OpenCV thing to read what they are

if the game is DOTA or LoL there might even be an API you can access to get that data

DOTA doesn’t need video analysis, it logs (and, I think, transmits) information about individual positions and state

the replay is not a video recording, it’s a board-resync

(mind you, I haven’t actually inspected the data stream, just played a long time ago, and used the various stats / gameplay analysis services)