YouTube Transcript:
Real time image data collection using python | Custom Object Detection | TensorFlow

Skip watching entire videos - get the full transcript, search for keywords, and copy with one click.

Video Transcript

View:

this is Dr sadik and today we are

discussing how to collect imagery data

using Python opencv and your webcam as a

data scientist you need to have a very

huge amount of data for object detection

and pubg detection model implementation

in a specific application areas such as

driverless car emotion detection sign

language implementation and

optimizations if you are a deep learning

researcher and you are a data scientist

then today's topic is very important for

you you can easily collect imagery data

about different subjects in about

different classes like if you want to

work in emotions so you can collect data

about different emotions if you are

working in object detection in action

recognitions then you can easily collect

data at your home at your office in

through from your students and they can

give you different poses different

gestures different actions and they can

they can pretend to have different

emotion as well so it's very easy to

collect data about different emotions

and about different actions uh using

python code you can collect enough

amount of data for your model and you

can initiate object detection model at

your home so I will directly jump to the

to the coding because we already know

how to install tensorflow and how to

install different dependencies of

pythons and object detection

so if you are new here and you don't

know how to install tensorflow in other

dependencies then you can find my

lectures how to install tensorflow and

how to install Cuda in other

dependencies link is in the description

below so in this lecture we will cover

three types of actions one is like this

one like for example this is for like

this is for dislike and this is for no

this is for no so we will cover these

three types of actions in this lecture

so let's have a look

so let's start

create a folder with imagery data new

folder imagery data

so now go to the Anaconda prompt and I

will activate uh OTF obj

I have already created this environment

and installed the dependencies of FG

detection like Cuda Cubana and

tensorflow object detection so if you

want to learn about that I have already

provided the links in the description so

you can follow all these lectures

so I will directly go I will activate

this TF obj

so go to the D

and in the D you will have to open the

jupyter notebook

Jupiter notebook

so now go to the imagery data and in the

imagery data create a python like an

interpreter file

new python file

so now we are in the D and in the D we

are we we are in the folder images data

we have created a jupyter notebook file

and now import dependencies import

CV2

import

OS import

time import

uuid

so this is actually opencv

OS is for the file path

and the time we use this dependency for

to give some bread between the different

classes like for example if we have

three classes one is like dislike and no

so if if the system collected images for

like so time break a time is also for

time taken is also for the the time

between the two two images and lastly we

will use uuid to name our image file so

let's run this code

okay it's successfully implemented let's

move to the next step so next we will we

will have we will give a path to the to

that particular folder and in that

folder imagery imagery data we will

create another folder which uh name is

image folder okay so here you can see

this is the imagery data and in that

image data we don't have any folder here

you can see uh imagery data

and in the imagery data we will create

another folder and its name should be

image folder okay

so let's run this code

and now let's move to the next step and

in the next step we will provide labels

like we have three labels like this like

No And we want 10 picture for every

level like we are giving this this

command that collect 10 picture for

every every levels so let's run this

code and move to the next step so now we

have three labels like dislike and no

and we need 10 uh 10 images per level

and here in the next segment we are

giving birth and we are we are trying to

create a directory and next this one

will access your webcam and my webcam

device name is it's zero it's

independent your system then what what

what is your your it can be one two or

something like that so here this code

will will access our webcam and it will

initiate our webcam for my device it's

zero and it depends on your system

Network number uh should fit in your

device so here in the next step it will

print that we are collecting images for

a specific level it depends that what

label is like for every level it will

give you a message a print it will give

you a print message so next is time that

slip so it will give you five seconds to

to have another pose

so this is important because you can

give different poses of a different

action like if you are collecting images

for emotions so you can give different

poses different gestures about different

emotions and if you are giving some

different action like if you are working

for sign language then you can give

different actions as well so next we are

going to collect images and this code

will will give us the particular uh

objective we can we can achieve the

particular objective from this

particular code and here you can see the

total number of images and the total

number of images is 10 so we want 10

images per label this particular command

will give us a capture and we are most

interested in in the frame the next this

this is the entire path here we will

join the path here you can see image

image folder here we will give the this

particular path and it will create

and also label for because we want to we

want different folder for each label

because it will create different level

for like three it will create three

labels lying dislike and no and in each

level it will uh it will save the

particular actions like if we are giving

action for the uh for the like so it

will save it in the uh in in that in the

like folder and if we give a dislike so

it will save it in the dislike folder

and same is the No it will uh definitely

save it in the jpga format and it will

pass it through specific string which is

which means that it should be unique

here the CV2 will write that particular

image uh to that particular folder and

lastly it will show us on the screen as

well and lastly we will go to the uh it

will be the ability to second time for

sleep like between the two images and

lastly if there is any error so it will

go directly to break that particular uh

to stop that particular uh compilation

so let's run the code and see what is

going to happen

so there is an error so now let's run it

again uh I think it was uh because I

haven't initiated again because the uh

because the Canada was dead so whenever

your Canal is dead and you uh you

initiate it uh Jupiter notebooks so you

must

initiate the the all this code like

python image image path labels and next

then you can jump to this this

particular segments so let's start from

this run

okay next

and

finally

so let's wait for it it will take a few

few seconds

yeah it's collecting here you can see

collecting images for like so let's see

let's give pause for like

okay this is for like

here you can see

it will collect 10 images actually

so let's see

I'm giving different poses

now it will collect for dislike so I'm

giving D for uh

action for dislike so let's see

and the third folder will be no

okay now it's time to no collecting

images for no okay

so you can give different poses like we

we have done this for three

three classes and you can give so it's

all done so let's let's check and have a

look at

in the folder so it's already created

here you can see imagery

you can see imagery and in the imagery

we can have it's created we have created

for the image and in the image is

created image folder okay just like this

you can see here

here you can see it's created the image

folder like imagery and in the image we

have images and in the image folder uh

it created image folder here you can see

so we are in the image folder so for

each of the class it created different

folder like for the dislike it's

collected different images here you can

see it's

collected for dislike and

here is collected for like you can see

it's collected for likes and it's

collected for no actually we demanded 10

pictures per label so that's why it's

collected only 10 picture per label if

you want hundreds of thousands of

pictures so you can you you can you can

have very long experiment for that so

this is all about how to how to collect

image data using Python opencv and your

webcam so in the next slide we will

label the particular actions and in the

next lecture you will learn how to label

that particular images using python

coding and how to detect that that

action in real time thank you so much

Click on any text or timestamp to jump to that moment in the video

Most transcripts ready in under 5 seconds

One-Click Copy125+ LanguagesSearch ContentJump to Timestamps

Paste YouTube URL

Enter any YouTube video link to get the full transcript

Most transcripts ready in under 5 seconds

Get Our Chrome Extension

Get transcripts instantly without leaving YouTube. Install our Chrome extension for one-click access to any video's transcript directly on the watch page.

Add to Chrome — Free

Works with YouTube, Coursera, Udemy and more educational platforms

Get Instant Transcripts: Just Edit the Domain in Your Address Bar!

YouTube

←

→

↻

https://www.youtube.com/watch?v=UF8uR6Z6KLc

YoutubeToText

←

→

↻

https://youtubetotext.net/watch?v=UF8uR6Z6KLc

YouTube Transcript:Real time image data collection using python | Custom Object Detection | TensorFlow

Video Transcript

Paste YouTube URL

Transcript Extraction Form

Get Our Chrome Extension

Get Instant Transcripts: Just Edit the Domain in Your Address Bar!

YouTube Transcript:
Real time image data collection using python | Custom Object Detection | TensorFlow