 have any idea of what AI is? Yeah so what is AI? AI is some technique that will all machine to work like human right? Yeah but saying so it is not not only a robot that is a replica of human. It may it may arise from a basic narrowed AI that is expertise in some certain field to the artificial general intelligence. So there are different techniques that can be applied to to have artificial intelligence. One of the technique is machine learning. Machine learning is the technique where you make your make your model launch certain patterns from the feature extracted from the data. So in machine learning you need a data that you will actually make your machine learn from the data. You like to extract certain patterns from the features that are extracted from the data and then there is something called as deep learning that will use a complex network structure to make your model learn from launch certain patterns from the features. There are other terms like reinforcement learning and deep reinforcement learning too which will not be covered in this section. So as I've already told machine learning is a model or it is some mathematical algorithm that will take a data right? So it will take a data and tries to learn something from it. Tries to learn certain patterns from the data but before reading the reading the data on this one you will have to pre- process that data so that certain features are extracted and the model will launch certain patterns from it. So we do have two types of machine learning generally supervised learning and unsupervised learning. In supervised learning you will provide the output yourself which is called annotations along with the data. So the model will previously know that what it should actually get as the output. It will have input and the output yourself but that mathematical formula will also calculate the output. So you will have two output rights. Then you will compare that two output and find something called loss function. Model will understand the differences between what it have got and what it actually should have got. But in unsupervised learning what the model will do is take the data, just learn the data it will not know the output previously and try to figure out those certain patterns from those data and it clusterized those similar features into one clusters. So you can see there those apples are clusterized in one cluster and other banana are clusterized in another cluster. So it will be done by the model yourself. So let us look at the example spam classification. Almost all of us have used emails, Gmail right. There will be something called as spam, the emails which spam will be classified separately right. So that is usually using the machine learning. But the thing is the data over here is mail. Mail season is now what you have to do is something called as natural language preprocessing. You have to have the basic knowledge of data science for that. So for using machine learning you have to have the basic knowledge of data science, email preprocessing or that or natural language preprocessing feature engineering. You have to understand these subjects before. So here the initially there are incoming mails right and then the mails are analyzed. You will do it with some algorithms and then once the mails are analyzed the extra features are extracted from those mails only and then those features are only fit into the models. You won't fit the noisy data into the models and the machine learning model will do its tasks and perform performance work. Here the work is to classify it as spam or not spam. So let us look at the general application that you guys have used in a day to day base. So almost all of us know about the face recognition. You guys have used face lock right. Isn't it? So what does it use? It is a machine learning model isn't it? What is the data for that? Images isn't it? So what will happen is it will take image, it will recognize image right. For that you will have to feed certain training data set to the model before training for the training process. It will feed the images, it will perform its tasks and then while you just unlock your mobile with face it will figure out where your eyes are, where your nose where this basic structure of your face is isn't it? So this is the day to day application of machine learning. You guys also have used mails that spam classification yourself is an application of machine learning. You guys have also you guys also have used Siri isn't it? Siri that's speech recognition application of machine learning and then you guys also also having got just something like the auto there is the auto-generated word in like docs and what does it use? It is machine learning. What is does before machine learning is a natural language preprocessing. So there are many applications ranging from medical diagnosis to basic face lock to everything. We do use machine learning apps like in a daily base not app but model. So another topic is deep learning. As I've already told deep learning is self machine learning right but it will be using something really complicated structure like a network that is inspired from human or biological neuron nervous system. So you have neurons in every layers then each neurons are connected to the neurons in another level right. So here is neurons will have something like a machine learning model which I won't be like explaining in deep for now but if you guys want to know you can just ask me later. So let us try to connect machine learning and old press now. We do have few machine learning plugins that are that are in old press but still we do not have as much as like those plugins that are usually used in other section right. We don't have as many but let us see one plugin called Imagine. So there is this plugin called Imagine in old press that will generate the image as per the text you write. If you write a dog over there then this plugin will create the image of dog for you. So if you are a content writer it will be really useful for you isn't it. You want you can create the image as per the word or text you write or the sentence you write. So what it is actually doing is it is generated so it is inspired from a dolly generation generation algorithm that uses our diffusion algorithm. So what diffusion algorithm will do is add a noises in the image will add Gaussian noise in the image and then try to denoise those noise images. There is another plugin called the Bertha AI too and this Bertha AI will generate the content for you. If you write the topic then it will generate content for you. It will generate some words for you but in the free version you will have only you can generate only thousand words and in Imagine also you can only generate 40 images for free version. Now let's just look at the demo. So here is the example how you can use the imagine plugin. So here I've written a dog in the forest eating cake so it's generating that image for me. It's generating the image of dog and the forest and the cake for me. So what it's doing is as I've already told it was trained with the data with the images with the normal images that was fed into the model and then it denoised it added Gaussian noise to it and then it denoised it again to generate the artificial images. So as for the image style and artistic style that we just feed into this imagine plugin it will generate images for you. So let us look at the how we can actually generate model. Let us look at how we can generate more train the model using some languages. So I'm using Kaggle over here. Kaggle provides you GPU it also provides your platform to run the model and I'm using Python but we can just run the model using PHP or Java or R but Python have actually a large number of frame of large number of it provides you a lot of architecture models and everything so I'm using Python. So I've imported libraries over there and then I've loaded the data as I've already told my for machine learning data is the most right. So I've used some pre-processing and then I've extracted the feature with the base that will be done by the by this we have something called Mobile Lead V2 that will be using our CNN architecture which is deep neural network which is inspired from the deep neural network and it will just extract something some features from those images. Then I've defined some function with pre-processing and the model then I've compiled the model and then CSVlogger keeps the data of accuracy and every matrix you provide. Then you can just plot with a this a math plot and see if the accuracy is increasing and the laws are decreasing or not or different auger matrix you can see in the graph then then previously I use a base base feature extractor right so I've on on free some layers and then again try to run it then I've added something called as call that but what I'm doing here for the for those who are like new to this I'm just running the model which is taking the images as the input and doing something like something like classifying the butterfly classes or types of butterfly into 75 classes that's what I'm doing with the model. So you're getting it right I've trained the model for those who are new I have just trained the model taken the input right now what I how can I connect it with the a psp or like old press is I have I can just download it I can download it and put a jib file right so I'm saving the model that I've trained just now so I have downloaded the model right now I what I can do is keep that model and make make something like api to connect the model and the old press right that can that was I can do so I'm just trying to over here I'm just trying to create our create api so I will use that saved model to to create create something like api so that I can use it in the old press with something like a plugin or any other integration so you guys remember that I downloaded the that model from the Kaggle right so I kept it in this model I keep that file into this model and then I write some code in a psp to use that model and predict predict the class or the type of the a butterfly for future use or to or to use that model so I'm using that model over there and I'm returning the type of the butterfly and the confidence level so I'm trying it I have like keep that in the hero to suffer for free and then I'm just trying to predict the type of that butterfly over there so it's predicting so I've fed the image and it's predicting African giant something class and with the confidence level 0.51 right so I can just you can now predict everything or every images over here like if that's related to the butterfly I'm just trying to classify that butterfly into different types I'm just trying to figure out the types of butterfly yes now I can I now I've created the api now I can connect the old press plugin and that model with that right so I'm just trying to create some a plugin that will connect that model through the api I just created now so I've just uh now I achieved that uh file and then and then I will just add plugin in the old press so I'm just trying to like uh add the plugin I'm doing it locally for now so plugin is activated now will be activated so I'm adding a post now you can choose the file or the image of butterflies right and what it will do what our model will do is classify the type of that butterflies in it so it has predicted African giant with 53.86 confidence so why I decided to uh show you this thing is because there are really less number of uh plugins and all for the machine learning and AI I just want you guys to I just want you guys to know that we can just create any plugin for as and we can do the machine learning part in this Kaggle things and we can connect it to the old press right there are so many applications but still there are really less number of plugins in the old press so it I thought it would be really inspiring for you to create some models that will be useful in the day-to-day life and you guys can upload it as a plugin in the old press