Before we jump into the code, let's try to understand the data set that we are using. So here the goal is to identify a given image, whether it contains cat, dog, or panda. So our data set consists of a total 3000 images. And there are 1000 images of a dog 1000 images of a cat, and 1000 images of panda. So this is a balanced in asset, where you have equal to equal distribution of images of each of these animals. And all of these images are organized into folder with the appropriate name.
These images are often obtained from Dr. Adrian Roseburg. So great, thank you to him. And I will show you how the folder looks like. Okay, so we're ready Here I have copy the code. So this is the data set. So the data set have a folder called animals.
And animals have three folders, right? And cats, dogs and Panda and said cats folder. There are thousand images of a cat. Okay. Similarly, inside dog folder, there are thousand images of dogs. Right.
And similarly, the panda folder, there are thousand images of panda. So this is our data set. And this data set will be given to you to practice on your own way you want and then So what we do is first we split this data set are already labeled. That means, we already know that these images are for dog, cat and panda. Okay, so first thing we do is we split the data into training it has tests. Usually, we say 75% of the images, we split into training data set as remaining 25% we spend testing, right?
Or it could be at 20 6040 whatever percentage you would like. Then what we do is we use a training data set we use to train the model. So the data set consists of an images as well as the labels for these images, which will be used for training the model, a label is nothing but whether the image contains cat the actual label with a cat, dog or pet And then Telstra said, what we do is the digital said we hide the labels, we only feed the test data, the images to the model, and ask it to predict the label of the model or tested it. And then we compare that label with the actual label that is going to sit as a test the test it as it has, and then we evaluate whether how good our model is performing in time, so scoring data set is available into the GitHub, I have uploaded the data set here.
All you need to do at this URL, you're here, right? And you that will take you to the TV repository, right. So the entire source code is available here. Right. So you our data sets that I have uploaded is available here. Right?
And the code is available here, right individual classes. The entry point is this ipython file. That's the file we will walk through as we navigate to the code, right. So all you need to do is you if you want to follow along, what you do is you could download this zip file on your machine, right anywhere you can download that you would like. So this will download the entire source code, as well as the data set and you just unzip it and then just start using it. Right.
That's how you use the source code.