#ot1-perplexing-regexing
1 messages · Page 553 of 1
That's not how it works
^
I have a penguin sliver n5000 with integrated intel uhd 605
Ah
Also now that ik that’s not how it’s work I won’t try kde
Xubuntu should give you a snappy experience
RAM is where most of your issues will come from
^
Yea I figured that
You can tweak Plasma to use less RAM, though
I'd recommend using xubuntu, lubuntu, or just using a WM such as openbox
Openbox will be the lightest
Is that easy to customizes
Openbox? It's very customizable
You can make it look however you want
All wm are, the questions is, how hard is it
Xubuntu uses xfce, and that's really customizable too
It's pretty moderate
You'll have to know how to edit config files, and know your way around Linux
But nothing expert level
I think that’s something I will do another time
Yeah, i wouldn't recommend openbox to a new user
But I do plan on use a wm, just not now
Il compared with lubuntu
Alright
There both customizable so it’s a matter of which is faster
Qt da best
ok
you can make a bootable hard drive with linux in it

we don't tolerate talking about hacking in this sever
you don't need to ask that in every channel, thankyou!
If you don't know what it is, how come you want to learn it?
just felt cool
lol
I do have a little bit of understanding about it
very little very very very little
just wanted to listen from you guys
is it related with Cyber security? or what
White hat hacking?
hu?
yes
Ok
oonga boonga im hackerman
White Hat hacking is like legal hacking , where you test other people's like, application/website for them , and find any sort of penetration points, and get paid
You should start with easy CTFs. Spoiler alert: offsec can be quiet boring though
Long long ago ,so long ago, no one know how long ago, I spoke on this server
Nice
i will beat all of you in hangman fight me cowards
ommggggg yes, hello!
define 'that dude'
Did you just fork it?
Because it's at 1k for me
yes because how could i pass the opportunity
LOL
.gh repo microsoft/vscode-python
Python extension for Visual Studio Code
Forked from DonJayamanne/pythonVSCode
https://api.github.com/repos/microsoft/vscode-python
ohhh
.gh repo DonJayamanne/pythonVSCode
This extension is now maintained in the Microsoft fork.
^
Noice
lmfao i have both vscode and vscode insiders running at once
@young shoal When is public static void main used in code?
entry point for a java program
So you do that when you're starting a new java program? I see
Every java program starts with a class definition and amain function definition , so the function would me something like this
// class definition
// . . .
public static void main(String args[]) {
// . . .
}
I think the argument is wrong though
Which argument?
String[] args
^
I'm pretty sure it's String args[] tho
weird
the two aren't always exactly equivalent
String[] a, b; would create two String arrays
String a[], b would create one String array and one String
yes
my only problem with java. once ppl learn it they make everything a class
exactly
Python is object oriented but you dont have to make everything be a class
Whenever a language is purely something, and the interpreter doesn't help it , you know it's bad
Yes
except, java doesn't
idk it sucks
lol
tbf
you could call python even more OO then java
because not everything is an object in java
Well , functions in Python are not Objects, in that matter lmao
?
functions are very much objects in python
in fact
i think we have a nice big tag about it too
!fao did i get right
Calling vs. Referencing functions
When assigning a new name to a function, storing it in a container, or passing it as an argument, a common mistake made is to call the function. Instead of getting the actual function, you'll get its return value.
In Python you can treat function names just like any other variable. Assume there was a function called now that returns the current time. If you did x = now(), the current time would be assigned to x, but if you did x = now, the function now itself would be assigned to x. x and now would both equally reference the function.
Examples
# assigning new name
def foo():
return 'bar'
def spam():
return 'eggs'
baz = foo
baz() # returns 'bar'
ham = spam
ham() # returns 'eggs'
# storing in container
import math
functions = [math.sqrt, math.factorial, math.log]
functions[0](25) # returns 5.0
# the above equivalent to math.sqrt(25)
# passing as argument
class C:
builtin_open = staticmethod(open)
# open function is passed
# to the staticmethod class
yep
@west zephyr :white_check_mark: Your eval job has completed with return code 0.
<class 'function'>
even a docstring is an object 😛
right
Aren't they all objects
i have an objection
splish splash your objection is objectively trash
i did not have an objection objected against you but you objectively trashed me
Eyy
This explains it nicely https://doc.rust-lang.org/rust-by-example/error/option_unwrap.html
Rust by Example (RBE) is a collection of runnable examples that illustrate various Rust concepts and standard libraries.
It's useful when you want to write stuff quickly
so its like some null handling thing?
Read the rust book, it's the best rust tutorial out there
basically if you have a Result<T, E> calling unwrap on it will give you T and error out if it was E, and on Option<E> it will give the value inside Some() and if it is None it will error out
I should go sleep, gn guys
gn
oh
unwrap is basically a quick way of unwrapping values inside a Result or Option
its not recommended in production unless you are absolutely sure that the Result or Option will have Ok or Some
TL; DR: unwrap if you're lazy like me
tbh I've not even used much Rust lately
Oh lol the unwrap thing in Rust is so similar to a function named "bind" in haskell , lol wow
So considering you have a value MyType 10 , you can extract that 10 and do whatever with it, using the bind function , it would look something like
main = do
boundValue <- MyType 10
print $ boundValue * 2 -- prints 20
umm
error: format argument must be a string literal
--> DataTypes.rs:6:14
|
6 | println!(string);
| ^^^^^^
|
sorry, just new to rust
what is this error for
println!("{}", string);
¯_(ツ)_/¯
rust has pointers?
yes
and raw pointers if you use unsafe block
and smart pointers
to communicate with C code
oh
yeah
btw what is &
reference
&variable will mean a reference to the memory where the variable is placed
it wont actually store the data of variable
you use this to make sure that a variable is not passed to some other block
ohh
yes
but how to find the data of variable?
how do you dereference?
*variable
but for using methods u dont need to dereference
u dont have to (*variable).some_method()
to get its method
variable.some_method() will work because rust automatically dereferences
fn main() {
let my_var = 45;
print_var(&my_var);
println!("{}", my_var); // if you dont give reference to function, you cant use my_var here again
}
fn print_var(var: &i32) {
println!("{}", var);
}```
read the owner ship chapter in the rust book it explains well
well its an integer so it will just copy
just an example 
yes
is it recommended to let rust automatically deref?
also you only need to deref mutable references when you want to change the value
yes
other than that rust handles it
nice
lol
should i use use std::io
io is io
tokio io
You may have forgotten closing a parentheses one day and are still writing in it.
-random redditor
wut
what the hell udemy?? I saw a course for 15 bucks, then I logged in and now its 129$
oops
ok
ok

hmmmm
e
*impossible quiz

yes
!user

Created: 4 years, 3 months and 11 days ago
Profile: @finite sierra
ID: 271586885346918400
Joined: 1 year, 8 minutes and 42 seconds ago
Roles: <@&542431903886606399>, <@&518565788744024082>, <@&463658397560995840>, <@&764802720779337729>, <@&267630620367257601>
Total: 1
Active: 0
joined here 1 year ago, today 
happy cakeday

Damn haskell is impure lmao
!user
You are not allowed to use that command here. Please use the #bot-commands channel instead.
\😢
I didn't know like what the hell
wdym
happy birthday(?)
yes happy 1 year birthday
?
You can literally do side effects in any function
just tell me your address bro
joe
ask your fatha when he comes back with the milk
corona.., no cake for you
bruh
who's joe
😩
maybe you can use the milk to make cake
how
summon avery
unsafePerformIO, it's literally unsafe
wot
cmon do the thing
candie's dad
lol
who's candie
yuri's sister
who's yuri
Haskell.addTo(unsafe_langs)?
candice's cousin
Like let's say, you wanna print in a function that's not IO :
launchTheNukes :: IO ()
launchTheNukes = putStrLn "BOOM!"
evilSucc :: Int -> Int
evilSucc x = unsafePerformIO $ do
launchTheNukes
pure (succ x)
Boom , evilSucc is not IO, but it still does an IO action , which is a side effect
who's candice
joe's sister
who's joe
candie's dad
who's candie
yuri's sister
It's safer than any imperative language, this is just a backdoor for the IO monad'
who's yuri
candice's cousin
who's candice
joe's sister
who's joe
stop
ok
i know deez
he has ligma
must have got it from that evilSucc
.
ligma balls
sure bro
lmao
bruh
😞
the joe family tree

lol
lmao
lol

Huh

Yes. It's ALL objects.
Even comments and keywords are stored in objects :)
All objects.
bruh
Joe Mama
i just
drank a not very strong beer
my mom tell me to
try it
ended up drinking the whole can
Sinner
ok done
this is kinda a stab in the dark but does anyone know at what temperatures most he3 fuses into he4 in stars?
yeah, I've looked all over google but I just can't find anything whatsoever
wikipedia lists the temperatures for each he3->he4 branch, but with literally 0 sources, and I can't find anything about the core temperatures of various classes anyways
Who does know Golang? I am interesting is there a point to study it for Web-development.
Its a backend, so sure
i need help
With what?
Bought some udemy course... the site is so garbage. always top reviews, but then it turns out to be yet another indian dude who just rattles down terms and acronyms without explaining anything...
I wonder if he himself even understands any of it.
which one?
"augmented startups"
ah
you know them? i guess the practical parts are better
nope idk them
oh
aha
AHAHAHAHA
how the fuck did i fuck up my windows path
oh
shit
i remember how
yeah that was a dumb decision
bruh indeed
Are you in a VM
I get an error with PowerShell in Windows Terminal
but it's a VM and so I have to open up cmd first
and then it works
cold water tastes different than warm water idc
wait wrong server
still though
they taste different
i think temperature affects taste yes
different brands of bottled water also taste different
@rough sapphire do you use vsc?
whats VSC?
oh for sure, they genuinely put different ingredients to water, some put more electrolytes or overall minerals to affect tastes
:blink:
OH
yeah i get that a lot, but nah its just a design i made for my online persona, "the trickster"
i prefer pycharm personally, but whatever you feel like is easier to work with
i havent used VSC myself but i have seen people code with it, i just think pycharm is more easy to the eye persay
You know the vertical line shown by default in PyCharm that indicates max length for a line for scripts to be PEP8 compliant? Well i know I can disable it, but how important is it to actually wrap lines to that length in the real world?
When I first installed PyCharm I was like....wtf is this line doing here lol
im not that advanced but it generally hasnt crossed my mind, i think youd be fine with or without it
it lowkey doesnt matter lol
As long as it stays around 120 chars you're fine imo
yeah give or take
and well, if you're crossing that limit it is possible that the structure of your code is going wrong
for example, if you're like 4 levels of nesting in, you might want to extract the inner code to a separate function
yeah even if this is a problem you can make you code more efficient
if you're making 200 character long list comprehensions, it is likely unreadable too
so yeah
the char count isn't a hard limit
but it's sure as hell good to stay under it
cool, thanks all
which one do you prefer?
let x = "42".parse::<u32>() or let x: u32 = "42".parse()
i usually end up typing the type-annotated one, but i think the first one is more readable
the second one
same
kubuntu
Why do u say that
cent is also good
Ik consider thos
why not
I was just ask for your reason on why you would recommend that
Or were u just saying that
@last mantle
.
Ok
CentOS is horrible for desktops
Literally no single person I know has ever even considered using CentOS as a desktop
I see it from time to time on servers, but it's still kind of horrible
ah i see.
alright. I didn't hate it on that time, but not sure, im not that much into changing OS a lot now a days, so you may be true.
CentOS just isn't used anymore
Even on servers, it's kind of bad
Most people will prefer Debian or Ubuntu
RHEL, even
Eh, love KDE Plasma
Qt FTW
Helloo people
goodbye , world
If you want to learn Korean....You can join this server
lol
😂
**The Developer Conference 2021 **conference: AI and Data tracks CFP ends 2nd May, please submit your talks via https://cfp-intl.thedevconf.com.br/ (see https://www.linkedin.com/feed/update/urn:li:activity:6793976069184618496/ for more details)
I doubt it's much different
!paste
ahaha I forgot how pretty Minecraft looks with shaders
Then of course... full screen in F1-mode
You use edge?
Yes, I use Edge
I do too
saem
but not using bing , use google on edge
Yes, as do I
also openSUSE and Arch both have given me a new appreciation for the colors green and blue
DEV channel
?
which dev channel?
that official?
yes
yes, it's official
aight thanks
lol
1
should be params = {order_id: order['orderId']}
~~ crossposting ~~
shotgunning
Hey @spice crescent!
It looks like you tried to attach a Python file - please use a code-pasting service such as https://paste.pythondiscord.com
You don't want it to be printed?
then just remove the console.log(r.data) i'm guessing
hello
hi
If I use a library licensed under GPL, can I use MIT for my project?
🦗
guys look I made something I don't understand
👏 yay!
I've got a headache
if you do not understand the Qt part of it, i can explain
C++ part , heh
eh C++ looks nice
i can understand what you have written
button.show() huh
try button.move(x=, y=)
ah im blocked :(
#include <QApplication>
#include <QPushButton>
int main(int argc, char *argv[]) {
QApplication a(argc, argv);
QPushButton button("Hello world!", nullptr);
button.resize(200, 100);
button.move(0, 0);
button.show();
return QApplication::exec();
}
show() is not needed imo
Ok so app just executes whatever you have made
It's not a widget per se
It doesn't show on screen
yes
Window, dialog, widget is where you have to place widgets etc
You can choose any
Note that QMainWindow has an inbuilt layout
And needs a central widget
So , better choose dialog
Ig in C++ its
QMainWindow root(parent=None)
Or whatever c uses
Oops make that QDialog
;-;
#include <QApplication>
#include <QPushButton>
#include <QDialog>
int main(int argc, char *argv[]) {
QApplication a(argc, argv);
QDialog root(nullptr);
root.showNormal();
QPushButton button("Hello world!", nullptr);
button.resize(200, 100);
button.move(0, 0);
// button.show();
return QApplication::exec();
}
yes and it shows a blank little window
Good
Note: it's better to make a separate class that inherits QDialog
oh boy
uh I understand what you've said so far
Goooood
class MainWindow: QDialog {
}
Yes if that's how c works
I believe that's how inheritance works
So how do you do the super thing
uh
If you want to understand things better use QtDesigner
shouldnt there be a proper tutorial
oof
😔
What's super do lol
its just easier to make them
I never delved too deep into OOP
for the tutorial guys
super calls the parent clas
You can execute parent methods in the child class
In py, super().init(parent) is used
For the QDialog class
class MainWindow: public QDialog {
typedef QDialog super;
}
``` I think
just use godot lol
godot is not as good as Qt 😎
Same with Qt
you guys, good afternoon, uh, in the beginning, did you guys struggle with loop "for"?
No
though in my case i just use SDL since i have gotten too used to it
i did yes
dude, wtf is "for c in range", wtf is "c"? I know it's a variable, but... uh, how can we use it on code, it seems to work in some codes, but, in others
just using them more made me understand them
like a list, its ok, to print it
I am confused
c is every element in the thing you are looping on
@latent scaffold go on
I've got... a window but no button
at first iteration its the first value of the thing you looping on, then after that the second.. etc
What's the supper for, again?
#include <QApplication>
#include <QPushButton>
#include <QDialog>
class MainWindow: public QDialog {
typedef QDialog super;
public:
MainWindow() {
QPushButton button("Hello world!", nullptr);
button.resize(200, 100);
button.move(0, 0);
button.show();
}
};
int main(int argc, char *argv[]) {
QApplication app(argc, argv);
MainWindow window;
window.showNormal();
return QApplication::exec();
}
yeah, I need to train more
@terse bough first lets move to #ot2-never-nester’s-nightmare
Oh, ok
¯_(ツ)_/¯
uhuh
Then layout.addWidget(button)
And
super.setLayout(layout)
layout being the instance of QVBoxLayout
No viable conversion from 'QPushButton' to 'QWidget *'
layout.addWidget(button)
I still don't know why that's happening
Learn Qt in py and then use on cpp
That's the best way
QPushButton is a widget
is C++ doing a::b() like Python doing a.b()?
if b() is a static method, yes
ah. that did nothing
#include <QApplication>
#include <QPushButton>
#include <QDialog>
#include <QVBoxLayout>
class MainWindow: public QDialog {
typedef QDialog super;
public:
MainWindow() {
QVBoxLayout layout;
QPushButton button("Hello world!", nullptr);
button.resize(200, 100);
button.move(0, 0);
button.show();
layout.addWidget(&button);
super::setLayout(&layout);
}
};
int main(int argc, char *argv[]) {
QApplication app(argc, argv);
MainWindow window;
window.showNormal();
return QApplication::exec();
}
example app in PyQt6
;-;
that sidebar looks sick
^
ikr
it moves ;)
my head hurts
relatble
very less
damn
its all QML
import QtQuick 2.15
import QtQuick.Window 2.15
import QtQuick.Controls.Material 2.15
import QtQuick.Controls 2.15
Window {
Material.theme: Material.Dark
id: window
width: 1080
height: 720
minimumHeight: 720
minimumWidth: 1080
visible: true
color: "#00000000"
Rectangle {
id: mainRect
color: "#323232"
anchors.fill: parent
anchors.rightMargin: 10
anchors.leftMargin: 10
anchors.bottomMargin: 10
anchors.topMargin: 10
Rectangle {
id: topRect
x: 0
y: 0
height: 66
color: "#1f1f1f"
anchors.left: parent.left
anchors.right: parent.right
anchors.top: parent.top
anchors.topMargin: 0
anchors.leftMargin: 0
anchors.rightMargin: 0```
hmm
this is mine
i think ill still stick to flutter
how do u handle button clicks @last mantle
and other events
intereting
flutter is nice, dart sucks
no
DISCORD IS CHANGING THEIR ICON
@low chasm sorry I forgot I don't use Ot0 anymore
SAY GOOD BYE TO BLURPLE
lmao
WTF this blew up here an obligatory plug check out my animations youtube @ firebuug if you like that kinda shit
Herbert my beloved
hi herbert
haha
noooooo
we are not discord by any means, so no
Just cause the blurple won't be the primary discord color anymore it seems
blurple is our primary colour
Are you sure that's not just your phone?
yes
Or maybe they uploaded a different picture to the store
see for yourself https://play.google.com/store/apps/details?id=com.discord
Idk man
Maybe their beta color is different
The same way canary color is different
Or smth
But Discord isn't changing from blurple
That I knowfor sure
changelog comes out on may 4th
Blurple is the actual name lmao
Cause I’ve seen it but I dunno how to make it black
In Discord?
Yea
Well sucks to suck
yeah it leaked earlier for all beta testers
mhm
It certainly glows more on dark theme than it does on light. I don't have beta but I see the new logo in Play Store
its ugly imo
Yep, toned down stuff looked way better
They could've adjusted saturation only a bit, but they messed up other parameters as well
i don't like it
It's just so much in your face. Especially since the logo colour is normally a colour used in the themes (eg send button) - if they change the theme as well, it would be too distracting :/
Which tutorial do you follow? Or docs
For base PyQt, real python has a very good tutorial on it
There is learnpyqt.com too
For that app specifically, you can do it manually but it takes time
So, you can use base PyQt with QtDesigner
Or if you want to make things simpler, use Pyside2 with QtQuick
Using QtCreator / QDesignStudio
Which uses QML for the UI, python for the backend
@tawdry jacinth
For QML, KDAB has a very good tutorial but it's in C++
Not that much of a problem tho
what editor are you using? is that the default color scheme?
You mean from the screenshot?
yeah, the code is so much more readable than my current setup
It's CLion with Materiel Atom Dark
and the background is an image of Castiel haha
nice! looks great, thank you
yep lol
cheese
what's a simple word for saying "stacked on top of each other"
layered
nice.. can you give me another one
I have boxes.. at most, one box can be stacked on another
busty
uh no
bruh
yes?
well rounded
best thing is, Qt comes with material theme, fusion theme etc by default
lmfao
I definitely gotta work on the layout more
TL; DR: Qt Creator is kind of cool
use wot
mhm but you haven't gotten to the bugs yet
no that's Qt Creator
Designer doesn't work for me tho
eh?
hm
you are using it..
lol
I can design using it
but the code won't work
like...
it won't generate code
well yeah
like it's bugged
that's what I'm using
its clearly different from what you are using
It looks different but it's literally Qt Creator
Qt Designer is different for me
im using the creator for .qml files
fine whatever
lol
Qt is a little annoying sometimes ngl
like
Idk how to add a border to things
border is usually a property than a widget
See how dumb it looks without a border ;-;
Even with like adding border color
hey you guys
border color wont do anything if the border size is near 0
increase the border size
find it out
why do you need a border tho
i dont think any app uses borders nowadays
because I want borders
i suggest you start with a Qt tutorial before jumping into Designer
:(
because you know how to arrange widgets
and will figure some other stuff out too
but when you resize the window, everything goes to shit
;-;
computer specs? it's usually fast for me
oh. then that explains alot
yeah ninja defaults to multiple cores i think
while make u have to specifiy
and i frogot to specify so i think its running on one core
bruh
Okay
so what's the kernel size
kernel_size: An integer or tuple/list of 2 integers, specifying the height and width of the 2D convolution window. Can be a single integer to specify the same value for all spatial dimensions.
I don't quite understand this 😅
?
i'll show an image hold on
kk
so this filter size is 3*3 or 3*3*1 if we consider channel as 1. kernel means this 3
kernel_size: An integer or tuple/list of 2 integers, specifying the height and width of the 2D convolution window. Can be a single integer to specify the same value for all spatial dimensions.
now read this
Ooooooh so it's the size of the filter bascially?
yes
???
some authors name them dimensions of filter
yeah. and you know that channel size is supposed to be same right?
The channel size should be the same as the image channel size, right?
yes
so lets say its rbg image
it must have size of
a*b so our filter can be z*z*3
z means 3 or 4 or whatever. kernel size basically.
i think it means the number of filters, we can apply a lot filters at one layer.
Yeah . But then , I have a b&w image , meaning it's just 1 dimensional . So that means my kernel size should be z * z * 1, or z * z?
its okay both ways. i mean now you understand it, both are fine.
and your kernel size is z. your filter dimensions are z*z*1
Thanks for the information, appreciate it
But also , what's a good kernel size for face recognition
So i have like 46-47 of each image , and I want to umm .. train it
3 by 3 is okay? 10 by 10? 20 by 20? Or should I just tweak it?
not sure. there is no thumb rule for number of filters.
3*3 is good i think
no worries
Hmm , ic . so (3, 3) . And the filters?
you can make filters big in later layers
It's the amount of filters so I can literally do whatever I want?
Or is that ... restrained too?
also i tell you what, if you have time, you can watch andrew ng's video, if you want playlist link i can give you. he literally started explaining from 0.
I know about feed forward and concurrent nns , so I know the basics. But ig I have to watch a video on convolutionals
look the thing is, yeah kinda. people have experimented a lot. if you put too much your model may learn something more so you may need to drop out neurons anyways.
Overfitting ig
Well , I guess just putting it on ... like ... 10 is alright?
it's basics face recognition
yeah you can try. personally speaking, i have more of theoritical knowledge, i have used keras for an NVIDIA event but i was using vgg16 and adding layers on that
for dog species recognition.
and vgg16 is already proven great in performance.
Ah
My laptop is not like the most performant laptop , it has medium-low graphics card , and decent CPU . But ig I can do at least 70% accuracy with some optimizations
how much memory?
gpu?
you can use colab if you want to use nice gpu. i think 11/12 gbs. cuda works on it.
yeah I've used colab , but this is somehow better in some cases
ig I can TRAIN in colab, TEST locally
But it's not a lot of images anyway, just like, 5 different faces
46-47 imgs each
yeah you can train in colab and save to drive
and then take in your pc
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 465.19.01 Driver Version: 460.32.03 CUDA Version: 11.2 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|===============================+======================+======================|
| 0 Tesla T4 Off | 00000000:00:04.0 Off | 0 |
| N/A 42C P8 10W / 70W | 3MiB / 15109MiB | 0% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
Oooo
it gives Tesla with 15gb
yeah i know!
which GPU is that
So
Basically
I need one conv2d, one flatten layer, one dense with activation, and max pooling?
how much images?
oh okay.
umm again im weak on practical aspect so i can't comment on that but i think you may need more than that.
more layers?
O...Oh ...
So where should I put it exactly? after the flatten? after the dense? after the max pooling?
before flatten
we put conv2d kinda always before flatten. once they're flatten they're flatten
Right
so yeah you can put after max pooling
So the conv2d applies the filter to the image itself?
also any need of max pooling?
I doubt actually
because its just 50*50
yeah
wait i'll just show my model for number recognition
yeah mostly does, but depends, if you add paddding then you can preserve size
Ooo
jesus this site
the keras site?
hold on this is loading
oh okay
no nvidia place where the notebooks are
sure
import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Activation, Conv2D
model = Sequential()
model.add(Conv2D(10, 3, input_shape=(50, 50)))
Here's my model so far, what should I make the input shape of the next Conv2D?
50*50? 45*45?
the simple logic is, as we increase layers,
we make filters big and their numbers less
Oh so we kinda make a batch filter
Hm
But the filters affect the image size, right?
That changes the input shape
And how would I change the input shape corresponding to the previous layer?
yeah they do. they kinda make it small i think.
Yeah, but how much?
depends on kernel size, stride and padding
3*3 kernel size, I haven't defined any padding
so it will make it small
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import (
Dense,
Conv2D,
MaxPool2D,
Flatten,
Dropout,
BatchNormalization,
)
model = Sequential()
model.add(Conv2D(75, (3, 3), strides=1, padding="same", activation="relu",
input_shape=(28, 28, 1)))
model.add(BatchNormalization())
model.add(MaxPool2D((2, 2), strides=2, padding="same"))
model.add(Conv2D(50, (3, 3), strides=1, padding="same", activation="relu"))
model.add(Dropout(0.2))
model.add(BatchNormalization())
model.add(MaxPool2D((2, 2), strides=2, padding="same"))
model.add(Conv2D(25, (3, 3), strides=1, padding="same", activation="relu"))
model.add(BatchNormalization())
model.add(MaxPool2D((2, 2), strides=2, padding="same"))
model.add(Flatten())
model.add(Dense(units=512, activation="relu"))
model.add(Dropout(0.3))
model.add(Dense(units=num_classes, activation="softmax"))
thats the model i used for ASL
Hmmm
So you have a conv2d with A LOT of filters apparently . then after some normalization and maxpooling another conv2d with ALOT of filters again, wow
oooo
Well , I don't know most of the layer types anyway, lol . Just conv2d, maxpooling and dense, a little bit flatten . But ig I can still do stuff with them, can't i?
Why did I say "does' lol
you can! but i personally prefer learning them before. you know dropping out?
n...no 😅
Dropout is a technique for preventing overfitting. Dropout randomly selects a subset of neurons and turns them off, so that they do not participate in forward or backward propagation in that particular pass. This helps to make sure that the network is robust and redundant, and does not rely on any one area to come up with answers.
Oh so basically anti-overfit
But I would need more epochs then
because it selects randomly at first ig
yeah.
i'll say try to make basic models first and see out it goes, may be more number of epoch may help.
Well , telling the difference between faces is a basic model isn't it? Usually the first projects are between a cat and dog, arent they?
and if you want, you can also augment vgg.
ooo
no that kinda uses a lot layers too. even most basic models atleast used to use 2-3 layers
and then AlexNet came and deep learning came in pictures with hella layers
noi noi, its aight, chill it out!
you'll learn in no time.
i dont think i could answer this much before 15 days, its all thanks to andrew.
you can see this video for famous models
Take the Deep Learning Specialization: http://bit.ly/2VMlo09
Check out all our courses: https://www.deeplearning.ai
Subscribe to The Batch, our weekly newsletter: https://www.deeplearning.ai/thebatch
Follow us:
Twitter: https://twitter.com/deeplearningai_
Facebook: https://www.facebook.com/deeplearningHQ/
Linkedin: https://www.linkedin.com/com...
Ooooo
I've also watched Luis Serrano's NN series, well not all of it , but ig it was clear
Oooo
thats LeNet, even this has 2 convNets.
and 2 FCs at the end
Damn
and thats i think before 2000s
wow
