Machine Learning Projects You NEVER Knew Existed

so you're looking for machine learning projects well in this video we're going to look at machine learning projects at three different levels we'll start off with beginner level projects we'll then take a look at some intermediate projects and last but not least we're going to cap it off and we're going to take a look at some advanced projects that you can use to get started with your machine learning deep learning or data science career let's do it [Music] right so i am super passionate about project i think it is the key thing that you should do regardless of what type of development or coding you're doing is going to be the thing that allows you to go from zero to 100 real quick particularly when it comes to advancing and building up your skills within a particular field now this doesn't matter whether or not you're doing web development or you're doing deep learning doing projects is going to help you accelerate a ton faster like i mean there is going to be nothing else that helps you progress further than actually doing projects now in this video we're actually going to take a look at projects at three different levels and they're going to be in reference to ryan reynolds different characters so we're going to start off and take a look at some beginner level projects and really these think of these as ryan reynolds in the green lantern so i mean they're good but they're not crazy great we're then going to take it one step further and we're going to go into intermediate territory so this is getting into deadpool level projects so you're going to need a little bit more skill and it's going to be a little bit more challenging but that means it's going to be great and then we're going to take it all the way to the max and we're going to take a look at some advanced level projects and to be honest i think when we're looking at these level projects we're looking at the dude character from free guy am i right let me know in the comments so what we're going to do is we're going to take a look at each one of these projects and i'm going to link to different examples that you can actually give a crap in the description below some of them are going to be available on this channel other ones are going to be available at other places but i'll link all of them below you can give them a crack and actually start adding them to your github portfolio because that is the second most important thing once you've done these projects you really got to show people what you're doing particularly in order to get value from it for your career so remember that once you've done the project make sure you share it and tell people about it all right without further ado let's start with our beginner or green lantern level projects so we've got three beginner level projects that i'm going to recommend you take a look at when you're just getting started now a large number of these beginner level projects are going to be to do with working with tabular based data so think about working with csvs or working with excel files so the first project is learning how to predict churn this is really important for most modern businesses because it's all to do with trying to predict whether or not a customer is likely to leave a business or not now this is based on the fact that it costs a lot more to attract a new customer than to try to retain a customer so this is why businesses optimize and really focus on ensuring they minimize how many customers they lose because it's a lot more cost efficient and ideally you want to ensure that you're delivering good customer value so for this project you're typically going to be working with a tabular-based data set in excel or csv and you'll have a whole bunch of customer features and what you're trying to do is predict a1 whether or not a customer is churned or a zero whether or not they're likely to stay with the business but that is the first project to take a look at the second project is a little bit different but again still has a large proportion of value when it comes to businesses and it's all to do with forecasting a business's sales forecasting sales allows a business to better estimate what's likely to happen in the future and have a better idea of their cash flow because cash is king in most businesses so when you go to forecast sales ideally what you'll be doing is you'll be passing through a number of different product like features whether or not you're running promotions whether or not there's discounts on at a particular point in time as well as different components when it comes to seasonality so what proportion of the time of the year is it is it easter is it christmas those tend to be really really important factors and what you'll be doing is you'll be outputting a continuous value using a regression model so this is all to do with predicting a continuous value now typically you can do this with a psychic learn model and there's a whole bunch of algorithms that help support that the third project that i'd suggest you take a look at as a beginner is sentiment analysis and specifically take a look at working with the twitter api so this definitely allows you to extend out your skill set a little bit further because it gets you interactively working with different data sources that are out there the working of the twitter api is going to expose you with how to authenticate to different apis and it's also going to teach you how to process different types of data and process json so that's going to be the result of the data that you actually get back now there's a whole bunch of different algorithms that support sentiment analysis but i'd highly recommend you take a look at the nltk library which is available through python or potentially take a look at extending yourself out a little bit and seeing if you can do something yourself and build up your own sentiment like model so what's going to be happening when you go to predict or produce one of these models is you'll be passing through your tweets as a series of tokens and you'll be outputting ideally a value between zero and one so one being positive sentiment zero being negative sentiment certain models break this out a little bit different into uh objectivity and subjectivity as well so those are our three beginner level projects so predicting churn forecasting sales and twitter sentiment prediction so this brings us to our intermediate level projects or really starting to get into deadpool territory so the first intermediate level project is really going to start stretching our boundaries and get us into computer vision now this particular project is actually available on the channel and it is all to do with automatic number plate detection so if you've ever walked into a new style supermarket and you've actually gone and haven't actually received a ticket this is because those supermarkets are starting to use something called automatic number plate recognition a camera is scanning your car it's extracting the number plate and it's actually extracting the text from that number plate to be able to check your registration against their ticket-based database so the way that you actually do this is using computer vision so first you actually use an object detector to determine where the number plate is and then you can actually use a optical character recognition algorithm to actually extract the text out of that plate from there really the world's yours so you can start to do a lot with it and use that number plate text to go on out there so our second intermediate level project is actually using a really advanced model in the background but actually allows you to get started with a little bit less difficulty so this is all to do with text generation using transformer models now transformer models are completely shaking up the deep learning world because one they're a lot faster and true it requires a lot less training and three they're a lot more accurate than traditional deep learning like techniques that being said you can actually get started with them relatively quickly using a library called hugging face the cool thing about this is that you can actually pass through some text and actually have it generate new text for you so say for example you wanted to generate a summarization based model or say for example you wanted to get your machine learning model to write a poem the hugging face transformers models actually allow you to do a whole bunch of really really cool stuff and start exposing you to the world of natural language processing so there's a whole heap of opportunity there if you're particularly interested in working with natural texts our third intermediate level project is starting to get a little bit more active and this is all to do with exercise correction using key point or landmark detection the cool thing about the deep learning world is a large number of organizations share their open source models out there with the community so you can actually get started now mediapipe is actually one of those libraries which actually gives you some really really sophisticated deep learning libraries that you can actually use to go on ahead and start building out your own models now i've actually built up one of these models on the channel so you can actually go on ahead and actually build a bicep curl tracker now again it probably doesn't have too much business value but it is a really really cool project which actually has a really visual component you can actually see the impact of your machine learning model counting bicep curls the fourth intermediate level project is common toxicity classification so this is again going down the natural language path so you will be passing through a series of tokens or really a sentence and you're trying to determine different levels of toxicity so when you think about uh how facebook actually processes comments and tries to flag comments which maybe aren't conforming to its expectations or um comments which are a little bit mean or bullying like this is a great example of where you can use this and again this has a bunch of applications could implement it on your own web apps you could um scan different blocks of text if you're working with schools again something super important when it comes to minimizing the impact of bullying now the way you'd actually go about this is you again you'd convert your sentences into tokens but you might actually take a slightly more sophisticated method and rather just replicating what you did for twitter you could actually build your own deep learning model to ascertain whether or not or different levels of toxicity so i know there's a couple of data sets out there that actually break toxicity down into different components and again i'll link to that data set in the description below all right so those are our four intermediate level project so we had automatic number plate detection x generation using transformers exercise correction using media pipe and key point detection as well as comment toxicity classification now this brings us to our dude or advanced level projects so there's going to be five projects that i'm actually going to reference here and again the majority of these are going to be available on the channel so you can actually give them a crack yourself the first one is one that i'm actively working on right now and building up from scratch and that is image super resolution i think this is so so cool because you can actually pass through a low resolution small image and train a machine learning model to actually enhance it so it's almost like uh having your very own sort of jarvis-like technique so you can pass through an image and go enhance and it's gonna return a high-resolution image the way it does this is using a generative adversarial neural network so what we do is we take high-resolution images and then we actually turn them to crap so we actually blur them and rescale them so that they're actually smaller and then we pass it to a specific type of neural network called again and we train that neural network to be able to enhance it to produce a higher resolution image again you could actually implement this on your website if you had some people that wanted to boost the quality of their images this is another great example the second advanced project is leveraging or building your very own game machine learning model using reinforcement learning so reinforcement learning is something i'm super passionate about and there's a whole heap of different applications for it out there that you can actually get started with but really it's all to do with teaching a machine learning model how to interact in order to maximize its outcome so this is a great application when it comes to playing games so if you wanted to say for example teach a machine learning model how to play flappy bird or how to play space invaders this is the exact space where you can go and deploy a reinforcement learning model and it's one that we've actually done on this channel so we actually taught a reinforcement learning agent to one land a spaceship we also taught it how to play space invaders we taught out how to optimally change our shower temperature to give us the best possible shower so there's a whole bunch of use cases when it comes to reinforcement learning and for practical projects the third advanced project is one which helps to bridge the gap between different cultures and that is neural machine translation so you can actually build machine learning models that are able to translate languages so natural text into a completely different language so there's a whole heap of opportunity when it comes to this but ideally what you'll be passing through is a sentence or string of tokens and your machine learning model will be sequentially generating new tokens to convert that into a different language so i don't believe we've done this one in the on the channel but i will link to this in the description below for some examples where you can actually get started with that the third advanced machine learning project that you can take a look at is action recognition so if you've ever seen object detection tutorials you can train models to be able to detect your phone and detect different objects but action recognition actually takes this one step further so rather than taking a single frame and asking a deep learning model to determine what objects are in that frame when we're performing action detection we're actually passing through a sequence of frames and going based on this sequence what is the action actually being performed so again it takes a little bit further now there's a whole bunch of use cases for this so say for example you wanted to perform sign language detections to determine what a particular person is assigning or if you wanted to actually use it to classify threat-based detection there are a whole bunch of different projects you can actually take a look at in terms of when it comes to performing action recognition and this brings us to our fifth and final advanced machine learning project that you should take a look at and that is neural style transfer so if you've ever seen a picasso or a monet style painting you know that they have very specific styles now mural style transfer allows you to take the style from a painting and overlay it onto another image so you could actually produce a picasso like photo of yourself and you can actually train again a generative adversarial neural network to be able to go and perform that style transfer on a number of images which again gives you the ability to be a little bit more creative when it comes to performing machine learning and on that note that is our fifth and final machine learning project that you should do in order to go and add to your portfolio so we took a look at five different ones so those were image super resolution using gans building gamer ai using reinforcement learning performing machine translation sports and action recognition using a whole bunch of other different use cases and last but not least neural style transfer using generative adversarial neural network and on that note that does wrap up all of our different machine learning projects that you can get started with so again we took a look at our beginner level projects or our green lantern style projects took a look at our deadpool intermediate projects and last but not least we took a look at our dude character advance project and on that note that about wraps it up thanks so much for tuning in guys hopefully you enjoyed this video if you did be sure to give it a big thumbs up hit subscribe and click that bell and let me know what you thought of this video i am trying out a bunch of different types of content i mean you've seen me do a couple of streams couple of shorts i figured i'd do a little bit of informational based stuff like this video let me know if you like it i don't know i'm just trying it out but who knows we'll see where it goes thanks again tuning in peace

Channel: Nicholas Renotte

Views: 4,612

Rating: 4.9492388 out of 5

Keywords: machine learning projects, machine learning, machine learning projects in python, machine learning projects for beginners, ml project ideas, ml projects for beginners, ai projects for beginners, ml project ideas for beginners, machine learning project

Id: sw3o0rAazMg

Length: 15min 19sec (919 seconds)

Published: Fri Sep 17 2021