Let me explain you the code structure. So this is our root directory where I have unzipped the code. Our main code inside the Canaan file. And our all code that we are going to be using this lecture is in Python notebook. Right. So before we go there, let's go through some common core that we will use for all image processing.
So there are two classes that we use. One is the data set loader. Right, and the other one is data set preprocessor. Right. So let's go through each class and understand what each class is doing. And then we can use this class directly into our IPython notebook.
So let's go to the data set loader class. I have opened that class here. Let me zoom in a little bit so that you understand. Okay, so this class is nothing but it imports NumPy See v2, and then less right? See v2 is the entire library that we used for processing image. So this data set loader class has a constructor, right?
And the constructor phase takes a preprocessor as input preprocessor is nothing but another class that does a modification on the image like pre resizing, and we will go through this class later. So for this discussion, just assume that there's a preprocessor class, it has a loader takes the input, and then it up, then it stores it internally, right? And then when it's loading, it applies that pre processor for each image. So this is the main function load function of the rest of the class where all the action is happening. The input to that is an image path, which is a directory consisting of all the images and then what was it just indicates the log level, what is how much log level you want, right? So we initialize the data and labels to the empty nest because this is what we want to construct, while we process the image.
So image path is nothing but a folder, right? So what we do is we enumerate through the folder, right? And then what we do is, we're just printing out some debug statements to for you to see what's an image path. Then we use open CV library CV to, to read the image from that image path, each path basically there are 3000 images, this loop will run 3000 times right, so for each image, it will read that image into an image object, right? And then it will read the label of the full name of the folder as a label, right? So if I go let me go back Back to the folder and show you how the images are.
Right so here are the images right. So the folder name is the image label. And inside each folder we have animals. So inside cats folder we have cats, dogs folder has dogs and so on. So what we are going to do is we are going to read all the images, then we are going to read the parent folder and make and put it as a label. Okay, that's our output, meaning that all the labels, all the images in the directory are cats.
So it extracts the image and extract the label from the folder name and then internally and then it applies a preprocessor. For the each image one of the purposes that we will be studying is the use of the preprocessor that resizes each image into 32 by 32 pixels, right? So it applies that preprocessor and then it applies the ads The image to a data list, and then label to a label list and then it continues, continues again to the next image. So the loop will continue for all 3000 images. And then at the end of the loop, we'll construct the data which consists of all the images, right, and the labels, which consists of all the folder names. So this data, basically when it will read through an image, Siri to library, it's going to read each pixel density, right?
And then we will see what will be the size of the image in the later course. Okay, okay. I didn't go to return those lists. So that's about dataset loader class. So the purpose of it is it loader classes, given an image path, read all the images, read all the labels and create a data set which a data set a features feature dataset, where is the actual image is there and then I will data set with labels. That's all it is.
A learner does