3. Data Annotation & Preparation Yolo v7 | Object Detection | Computer Vision

hello everyone welcome to this video in this video I'll be showing you uh like how to do The annotation and how to prepare your data set for this year of V7 so uh to prepare your annotation file okay you need a tool called label mg I think you would be familiar with this level mg if you have work in object detection before okay so this is the tool actually uh I will give everything in your researchers section okay so that you can download and get started with so what you need to do uh you just need to click on this level mg tool okay you don't have to install anything so it will open up a UI for you okay so see this is the Y of this level mg so now uh here uh you need to open your directory so if I show you the images so guys if you see this is my images okay I will be using this sign language detection images okay uh this image is actually collected uh so there actually will find some activity like hello and uh like uh I love you and no yes please okay there are six uh activity we'll get like if you do like that that means you are trying to say I love you if you do like that like uh so you are trying to say hello okay so if you are if you're doing like that so you are trying to say like please okay so that's how there are like six activity okay we'll get uh so now let's start doing The annotation because this is the raw images okay if I pass this images to my model okay it won't be able to learn anything so to make my model understand everything okay I need to annotate the data so that's why I will be using this annotation tool okay so first of all you need to open directory okay like why do you have kept your images so this is my images I have kept like all images folder so I will select the folder uh yeah select the folder okay so now uh it has opened my image here okay now what I need to do I need to uh change save directory because after doing The annotation okay where I want to save all the txt file okay so that folder I need to select so I will select the same folder so if you see like I selected this folder so I will select the same folder here uh select okay now here in the view section there is another option called Auto saving okay you can uh like uh mark this okay because it will auto save your email annotation file and not okay and here you can see at the left hand side there is the option called Pascal VOC okay so this is the Coco format uh annotation okay if you want to do in Coco format then you can save it in Pascal BSC like it will save as XML file uh but here I actually were using yellow uh V7 okay our official version yellow V7 so it actually always takes input as yellow format so I will change to Yellow Okay so so it will save as dxt5 okay now here you need to select create rectangle box and you need to start The annotation so this is my reason of Interest okay I want to detect so let's mark it as hello okay because this is the hello sign and okay okay so once it is done so I will just do the same okay now if I open my folder all image folder so here you will see a file will be created called hello like this this txt files okay so this is this was the images and with corresponding that this this was like our txt files okay now if I open the file so here you can see some coordinates points okay so what are the coordinate points you are getting here so if I show you here yeah so these four coordinates points actually it will sell okay and based on that actually it has leveled this hello as zero okay so this these These are the information we need to train our model so like that actually we have to do The annotation for all the images okay like presented in my folder if you see uh we have to annotate all the images so now let's pick another example like uh instead of hello I will pick another example like called I love you okay so now if I just select a rectangle box and that's why I will select and instead of hello I will give I love you okay I love you and I will do okay okay now if I save it now if I show you uh this folder so here another text if I would be created called I love you okay and these are some information and I love you has denoted with one okay this is the level they have given so that's actually we need to finish all the annotation I I have just shown you one to two images because uh just for the demonstration purpose okay but uh how many number of images you have in the folder you need to do The annotation for all the images okay so now when uh you have done The annotation okay so now what you need to do you need to prepare your photo structure okay for this yellow V7 so let me show you the further structure like how to do that foreign structure that's actually going to make like uh first of all you need to create a folder called uh you can give any name like I have I have named it like data underscore yellow V7 okay inside that you need to create uh some of the folder and files suppose first the first folder you need to create create like images okay the and then the second photo you need to create the labels and uh another file integrate called class classes dot names okay so see uh the extension would be classes dot names if I open that file now let me show you so guys you can see uh classes dot names okay this is the file extension you need to create and inside that you will be keeping all of your name level names okay the way you have leveled your images okay that name actually you need to pass here like hello I love you know please thanks yes okay I had six levels so I I have kept like that so once it is done then you need to create another two folder a two file called train train.txt and well.txt okay but make sure this uh files should be empty okay because we'll be utilizing this file later on okay now inside image folder you have to create two folder called train and validation okay inside train will be keeping all of your training images okay like all of your training images I have kept like I have kept around 120 images for training and I said validation you need to keep your validation images like I have kept 30 number of like images okay and insert levels with corresponding that images you need to keep the level file like that txt file you have done The annotation see uh All The annotation file you will get but make sure this txt file should be with respect to that images like inside train I have kept these images right and you see this is the name okay like the same name and the same txt file you need to keep inside your level level folder if you see like under this trend I've kept the same thing okay with corresponding that in the validation folder you need to keep all the validation txt file okay so once it is done uh so you are ready to go okay you have prepared this folder instruction and all okay so now what you need to do you need to make a zip okay so like you need to select all the file and click on that I am using the seven zipper key you can use other zipping cancer tool okay so I will do the zip operation I will make a zip so guys you can see it has been zipped now what I need to do uh I will upload this data in my GitHub okay so that I can utilize it later on so I will just open my GitHub here so guys this is my GitHub uh now suppose I want to upload this data in this repository so I will click on that okay now here actually what I will do I will just uh open my folder okay open my folder and here you just need to drag and drop okay just select and paste it here okay so what it will do it will upload this data set inside your GitHub and if you do now comment changes okay it will like upload this data set in your GitHub repository so I have already done so I will do the cancel and uh see I've already done this is the my data okay data uh underscore like yellow V7 okay so now you need to click on that and there is a button uh download okay now if you just click on this uh download like right click on this download so you will get uh some of the option okay here actually you need to select copy link address okay now if I paste this link in a new tab so let me show you if I press this link in a new tab and if I hit enter so now see it will start downloading this data okay so that is what I need okay so I will be uh using call command to download this data okay inside my uh Google collab so it yeah it was all about this data preparation and for the structure creation uh like uh video okay I think you have got it right now like how to do like uh this uh data annotation and how to prepare your folder structure so in our next video I'll be showing you like how to uh use this data set okay in our collab.book and how to download this data set and I'll be showing you like how to download a pretend model okay and I will show you like how to use this data for training our model so till and uh yeah guys this was all about from this video and I think you got it so thank you so much guys for watching this video and I will see you next time [Music]

Channel: DSwithBappy

Views: 12,223

Keywords: data science, machine learning, deep learning, python, computer vision, artificial intelligence, natural language processing, neural network, 100 days of ml, how to learn data science, dswithbappy, ds with bappy, bappy, 100 days of machine learning, মেশিন লার্নিং, ডাটা সায়েন্স, ডাটা সায়েন্স বাংলা, machine learning bangla, bangla ml tutorial, machine learning bangla course, machine learning bangla tutorial, data science bangla course, yolov7, object, object detection using yolov7

Id: e7VSOi2Yvng

Length: 8min 41sec (521 seconds)

Published: Mon Oct 17 2022