Hey my name is Felipe and welcome to my channel. In
this video I'm going to show you how to train an object detector on your own custom data
using YOLO V8 and Google colab. And this is the notebook we are going to use in order to train
this object detector and you can see this notebook is comprised of 1, 2, 3, 4, 5 steps. So in only
five steps you will have your own object detector up and running and this notebook will be available in
the github repository of today's tutorial. But now let me show you something else because we are going
to use Google colab but we're also going to use Google Drive and this is a directory I have
created in my Google Drive in order to work on this project and you can see that this directory
is located in my drive computer vision engineer object detection YOLO V8 Google colab, this is the directory
I am going to use in order to upload all the data and all the files we need in order to work on
this project so the first step you will need to take is to create a directory in your google drive.
Now let's continue and now let me show you the data we are going to use in order to train this
object detector and you can see that these are many many pictures of squirrels so this is the dataset
I am going to use, I am going to train an object detector to detect squirrels but obviously you
can use whatever dataset you want in order to train this object detector. Now let me show you something
which is very very very important and it's exactly how you need to prepare the data and how you
need to structure the data in order to train this object detector this is a very very very important
step so please pay attention if I go here you can see I have a directory which is called data this
is the root directory of my data and if I open this directory you can see I have two additional
directories one of them is called images and the other one is called labels if I open images
you can see I have two additional directories one of them is called train and the other one is
called val and if I open train this is where all my training images are located so these are the
images I am going to use in order to train the object detector if I go back and if I open the
val directory these are all the images I am going to use as validation data these are the images I
am going to use in order to validate the object detector so now let's go back and now let me
show you what's in this directory which is called labels you can see I also have two additional
directories one of them is also called train the other one is also called val and if I open
train these are all the labels all the annotations of the training data, of the training images, if I
open the val directory these are all the labels all the annotations of the validation data so
it's very very very important you structure the data exactly like this you need a root directory
then within this directory you have two additional directories one of them is called images another one
is called labels and it's very very very important these two directories are called exactly like this
one of them should be called images and the other one should be called labels it's very important
you name these directories exactly like this then within 'images' you will have two additional
directories one of them should be called train and the other one should be called val, it's very
important you name these directories exactly like this and then within these two directories is
where you will locate all of your data all the training data and all the validation data and
then for the labels it's exactly the same you need two directories one of them should be called
train and the other one should be called val it's very important you name these directories exactly
like this and then this is where you're going to locate all your labels all your annotations all
the training labels and all the validation labels so please pay attention because it's very very
very important you structure the data exactly like this otherwise it's not going to work
so please please please make sure you have structured data into exactly this format and once
you are sure everything complies with this format then you are ready to continue then the only thing
you need to do is to upload this data into your Google Drive so once everything complies with the
format I showed you the only thing you need to do is to take this directory and to drag and drop
this directory into your Google Drive and that that's pretty much all and then you just need to
wait until the uploading is completed and that's pretty much all. Now let me show you something else
which is the dataset I used in order to download all the images of all the squirrels I am going
to use in this video, all these images I showed you a few minutes ago and also all the annotations
I am going to use in order to train this object detector and it's the Google open images dataset
version 7 this is an amazing dataset, an amazing amazing dataset and I definitely recommend
you to take a look at this dataset because you are going to find many many many different
images many different categories many different type of annotations you have bounding boxes for
object detection you also have masks for semantic segmentation and this is an amazing datset, I have
used it many many times many times in my projects and this is exactly the dataset from which I have
downloaded all the images and all the annotations of squirrels so this is an amazing dataset and
I recommend you to take a look at this website and I also created a python script and a video
with very detailed instructions on how you can download data from this dataset and I tried super
super super hard to make this process super simple to make all this downloading process super simple
so by just following the instructions I provide in that video and by using that script you will be
able to create whatever object detection dataset you want you can just download images from all
the categories you want and you can just create your dataset in only a few minutes and this
is an amazing resource I created in order to help you to train your object detection models
and this is a resource which is available in my patreon so it's going to be available to all my
patreon supporters. Now let's continue, now let me show you how to execute all the steps in this
process and how to train this object detector and the only thing we need to do now is to execute
all the cells in this notebook one at a time and that's going to be pretty much all the first cell
is the one that's going to connect your Google drive with Google colab and this very very very
important because remember you will need to access all the data in your Google drive right... so you
need to select your account then you're going to scroll all the way down and you're going to click
allow and that's pretty pretty much all now Google Drive is mounted in your Google colab environment
that's pretty much all everything is ready so we can continue now the next step is defining this
variable which is root directory and this is the directory of your project in your Google
Drive, right, remember in my case I'm uploading all the files and all the data and so on into this
directory of my Google Drive so this is something you will need to change to the directory you have
created in your Google Drive remember in my case everything is located in my drive computer vision
engineer object detection yo8 Google colab and if I show you this variable this is what we have
over here my drive computer vision engineer and object detection yolov8 Google colab and then as everything
is mounted here on content gdrive you can see that everything is located in content gdrive
and then it's the location of my directory on my Google cab so please make sure you change this
variable to whatever location you have uploaded all of your data and whatever location you have
chosen for your directory in your Google drive this is very very important and once you have
changed this variable the only thing you need to do is to execute this cell and that's it then this
cell over here is going to install ultralytics which is the Python library you need in order to
train yolov8 on your custom dataset and this is very very very straightforward the only thing you
need to do is to execute this cell and that's it now if I scroll down we are almost almost almost
there you can see we have only two cells left and this one is very important... this is the one
we are going to use in order to do the training and in order to do so you can see that we need
another file which is google_colab_config.yaml now let me show you this file if I go to my file
system and I go up you can see I have a file which is google_colab_config.yaml and if I open this
file you can see that this is a yaml file with four keys: path, train, val and names, So it's very very
very important you change this value over here which is path to the location of your data in your
Google Drive you can see in my case the data is located in my drive computer vision engineer object
detection yolov8 google colab and then data and if I go here you can see everything is located in
my drive computer vision engineer object detection yolov8 google colab and then data but first I also
have these two... folders which are content gdrive and this is the location of my drive in my google
colab, right, so please pay attention and please make sure you have changed this value to whatever
location you have uploaded your data into, right, this is very very very important and then if you
have followed my instructions regarding how to create your dataset, how you structure your dataset,
then you don't need to change anything over here right this is the location of your training images
this is the location of your validation images and if you have followed my instructions
then you can just leave these two values... as they are right now right and it's going to work
just just fine just make sure you change this path this is very very very very very important and
I cannot stress this enough right this is very very important and then it's the names of all
the categories you are going to detect, in my case I'm only going to detect one category which
is squirrel so this is how it looks like for me but please change this value to whatever object
whatever category you are going to detect in your problem and if you are going to detect more than
one object the only thing you need to do is to do something like this right 1 is going to be your
category number one, then 2 is going to be the category number two this is going to be given
by your data right by your labels you are going to add all the categories you have in your data in
my case it's only one and its squirrel so once you have prepared this file what you need to do
is to take this file and you're going to upload this file into your Google drive right this is
exactly what you need to do the same way you did with this directory with your data directory you
need to do it with this file and that's pretty much all once everything is uploaded you are going
to get back here to your Google colab and now you can execute this cell so the only thing I'm
going to do now is to press enter and I'm going to execute this cell and that's pretty much all
that's pretty much all we need to do in order to train this object detector the only thing we
need to do now is to wait until this process is completed and this is going to depend on how many
images you're using it's going to depend on how many epochs you are going to train this model for,
in my case I'm going to train it for 20 epochs and I'm using something like 2,000 images if I'm
not mistaken so this is going to take a couple of hours I think this is going to take 1 or
3 hours and that's pretty much all okay now the training process is completed and let me show you
something let me show you all the output which was generated during the training process you can see
that we have a a lot of information and at the end it says results saved to and the directory where
the results were saved so now we need to execute the last cell in this notebook because we want to
copy all the content from this directory into our Google Drive now it's very important you change
this value over here which is the location of your directory in your Google Drive please remember to
change this value over here you can see that now it says my drive computer vision engineer object
detection yolov8 google colab which is the location of my directory in my Google Drive please remember to
change this value and once you changed this value the only thing you're going to do is to press
enter and that's pretty much all now let me show you my Google Drive this is the directory
I have just copied into my Google Drive you can see it says runs and if I open this directory it says
detect then train and these are all the results of the training process right so let me show you
now the only thing you need to do is to download this directory so I'm going back to... here and I
am going to... download like this okay now we need to wait a few minutes and that's pretty much
all in order to download this directory so this is going to download all the results you are going
to use in order to evaluate this training process and also it's going to download all the
weights of the model you trained which is very very very important let me show you super quickly if I
open this directory runs detect train then these are many many plots and many images you are going
to use in order to evaluate your training process let me show you super quickly this is perhaps
the most important of all of these plots and then this directory which is called weights this
is the directory which contains the model you have just trained so this is going to be pretty
much all for this tutorial this was a very very quick and a very straightforward tutorial on how
to train an object detector using YOLO V8 and if you have any questions regarding this process
I invite you to watch other of my previous videos where I show you the same process but in a much
much much more comprehensive way I show you the entire data preparation the entire process of how
to annotate your data how to evaluate the model you trained this previous video is much much much
more detailed so I definitely invite you to take a look at this video which I'm going to be posting
over there so this is going to be all for today my name is Felipe thank you so much for watching
this tutorial and see you on my next video.
