#can anyone help me with python machine learning?

73 messages · Page 1 of 1 (latest)

wintry lagoon
#

Explain problem here bro

#

!rule dm

astral ivyBOT
#
Direct Messages

Please keep discussion on the server. It helps you get responses more quickly from more people & with more viewpoints. It also helps protect you from scammers.

We recommend disabling DM's from all public servers.

urban kindle
#

i have to make a program that get a sound of person talking and it needs with background vs foreground to find the seconds the person talks
and it wants me to create an ML from the begining no ready dataframes.

jaunty hinge
#

What is "it"?

wintry lagoon
#

This sounds familiar?

#

Oh it's cause you asked this then left

urban kindle
#

i really need help.

jaunty hinge
urban kindle
urban kindle
# jaunty hinge .

i have to make a program that get a sound of person talking and it needs with background vs foreground to find the seconds the person talks
and it wants me to create an ML from the begining no ready dataframes.

jaunty hinge
#

I can read

#

"it wants me to [...]" -> who or what is "it"

urban kindle
#

recognise the seconds a person talks

jaunty hinge
#

just use whisperx lol

#

gives you the recognized texts with the timestamps for each sentence

urban kindle
#

i cant use ready i need to make my own and train it.

jaunty hinge
#

So presumably this is some sort of college assignment

#

and they presumably also gave you a dataset to work with?

urban kindle
#

i need to find a dataset

jaunty hinge
#

So you don't need to make your own model?

urban kindle
#

and use it to train the

urban kindle
jaunty hinge
urban kindle
#

i never use Ml and i need some help.

#

the problem is to make the Ml

jaunty hinge
urban kindle
#

the lesson is about sound not about Ml

#

the teacher just puts hard exercise

jaunty hinge
#

Sounds absolutely insane to have students do an ML model from scratch in some kind of audio class but whatever

urban kindle
#

i know

jaunty hinge
#

You just need to identify the timestamps?

#

What sounds more likely is that they want you to use some sort of concept you learned in the class (e.g. audio processing), seems highly unlikely that they expect you to just know ML 🤨

urban kindle
#

yeah like the person says the word "table" and i need to say he is saying it in 2,5 to 3 second of the sound.

jaunty hinge
#

You need to say "the person is speaking from 2.5 to 3 seconds" or do you also need the text content?

urban kindle
#

both

jaunty hinge
#

You're not training that from scratch

#

hell nah

urban kindle
#

what can i use.

#

?

jaunty hinge
#

I already recommended whisprx which does exactly what you need

urban kindle
#

You are tasked with implementing a system that segments a sentence into words, mandatorily using a background vs foreground classifier of your choice. Given a recording of a speaker, the system should return the time boundaries of the spoken words (in seconds). Additionally, you must provide an accompanying program that plays back the detected words. The number of words in the sentence is not known in advance, but you can assume that there is a small gap of silence between the words.

jaunty hinge
#

This is a completely different problem than the one you outlined lol

urban kindle
#

Attention!!!: You cannot use convolutional neural networks. The use of ready-made web services or APIs for speech recognition is not allowed. Transfer learning from pre-trained networks is also not allowed. Solutions that violate these rules will receive zero points.

urban kindle
#

and explain to me?

jaunty hinge
#

No

jaunty hinge
#

they tell you there are no background noises and that you have a pause between words

urban kindle
jaunty hinge
#

What

urban kindle
#

wait a bit

urban kindle
# jaunty hinge What

A) You are tasked with implementing a system that segments a sentence into words, mandatorily using a background vs foreground classifier of your choice. Given a recording of a speaker, the system should return the time boundaries of the spoken words (in seconds). Additionally, you must provide an accompanying program that plays back the detected words. The number of words in the sentence is not known in advance, but you may assume that there is a small silence interval between words.

You must implement and compare the performance of the following classifiers: Least Squares, SVM, RNN, and a 3-layer MLP (specify the number of neurons per layer). The comparison should be conducted as is typical for binary classification systems.

B) From the detected words, calculate the speaker's average fundamental frequency.

You must explain which data were used during the testing and training of the system. If they are your own, explain how you created them; if they are open source, explain how they were utilized.
Try to ensure that the system is speaker-independent and as robust as possible to variations in speaker characteristics.

jaunty hinge
#

See there you have it

#

they want you to implement a very basic thing and they tell you what to implement

urban kindle
#

can you give me a explanation of what i need to do?

jaunty hinge
urban kindle
#

can you give me a general explanatin

jaunty hinge
#

What did you do the entire semester 💀

jaunty hinge
urban kindle
#

the lesson is about sound

jaunty hinge
#

You probably ignored some of the requirements then? 💀

#

also never heard of an audio engineering class in a computer science degree

urban kindle
#

i really cant understand what i need to do.

jaunty hinge
#

Have you tried asking your professor for some help?

#

Cuz either you're lying or they're insanely incompetent lol

wintry lagoon
#

💀

#

Get his ass @jaunty hinge

urban kindle
#

now i understand what i need to do thanky you