Artificial Intelligence Full Course 2024 _ AI Tutorial For Beginners _ AI Full Course_ Intellipaat - Ep90

that okay maybe the snoutis round but pinecone ai use casealong with that the colorthat a pughas different as compared to a normalcat right so here we have only two uhfeature spaces but if youapply more number of features uh morenumber of convolution features on yourimage then every single feature rightmight or might not play an importantpart in making a classification butwhatever features will be present inyour model they will be contributing tocertain images right object detectionright so we will be doing objectdetection today as well becauseobviously CNN uh as I said 90 99% of theuse neural network models is aroundimage and video based data right so thedata that we have over here is uh Cipher10 data right now this data containsaround 60,000 uh samples and and all ofthese image samples are of the size 32cross 32 right now 32 cross 32 uh thesize is small the size is very small butthat is what we can work with in a 2hourclass right otherwise if you go aheadand if you start building a model onslightly larger images it will exceedthe modelbuilding so we are working with this butonce when we'll be done with this casestudy you will have the code and youwill have the tools to go ahead andapply this on any data right then youcan go ahead and you can download maybeimages from different domainsmaybe interest and then you can go aheadand you can try to create CNN modelswhich might be able to differentiatebetween the different kinds of imagesthat you have right now the diff the 10different classes that we have in thisdata are if you see over here areaeroplanes scars Birdsfrogs horses ships and trucks right nowyou can see that there are categorieswhich might be very similar in naturelike let's say if you take cats anddogs right and horses there arecategories which might be verydissimilar in nature as well if you seeAeroplane Aeroplane and birds rightAeroplane and deers so there is nocorrelation right so we'll see if we cancreate a see NN model which iseffectively able to classify all ofthese objects and right now see all ofthese objects are coming from differentuh different areas right it's not thatthat these are related the objects canbe totally random and we'll try to seeif the CNN model is able to capture thepatterns and is able to make effectivepredictions on these objects which arecoming from varying uh domains rightmulticlass right exactly so this will bea multiclass problem I have 10 classesright now problem statement is imageclassification classification using deepCNN in care detectionoperation right so let's proceed withthis let's import the libraries thatwill be needed uh so we will be usingnumpy M plot Li which is standard CV2 isa library for um visualizationand then from tensorflow docas we areimporting data sets we are importingmodels layers optimizers because all ofthese will be used uh from tensorflowdoc. preprocessing we are importingimage data generator right uh if youremember image data generator is afunction which will uh basically add newsamples of the same image right we arealso importing two new functions whichis model checkpoint and early stoppingso I'll talk about this once when wereach this uh these functions are beingused right and this all of this is notrequired we have just imported those sowe'll see whatever functions will bewhatever libraries will be used I willbe explaining about those right so IDefine my variables beforeand somethingthat generally you will do once when youhave a vision of the data right uh butuh right now becauseuh generally what we do all thevariables anything that you want tochange in your coderight that you generally Define towardsthestart so whenever you have to run thecode right you don't have to go insidethe different cells of the code tochange different variables right so makeit as a practice whenever you are codinguh so this will not happen in the firstiteration of the code because obviouslyin the first iteration of the codeyou're still writing the code you'restill thinking about the approach andall but once when your code is finalizedonce when you know that okay this iswhat I need to run and this is the codethat I need to run then what you need todo is that you need to bring up all thevariables that you say if somewhere youare defining aox right let's saysomewhere if you're defining the batsize right all of these things then youwill try to bring them up towards thefirst or second cell why now the nexttime let's say if you want to run thiscode all you to change the values if youneed to change the value of any variableyou just need to change the values of ofof a single cell that probably will bepresent somewhere on the top right andthen you can run the entire notebook anditshould right so that is why if you seewe are defining the bath size number ofclasses EPO and number of predictionsover here right these variables will beused at a laterstagenumber predictions anik is for theevaluatefunction once when we have built themodel and we want to evaluate it then itwill make predictions on a batch of20 we'll be using this today as wellright now once again we are using the Kdata so we are using data sets. siarright so if you see Keras as a libraryright it will have lots of datset right if you see I have these manydata sets I have Boston Housing Cipher10 Cipher 100 fashion Mist IMDb mnistrterright can go ahead and you can playaround with any of these data right whatwe are using over here we are usingthe cipher 10 data right soso if you want so let's say let's say ifyou do this right let's see what itreturns right so you can seethat let's go to this function as wellright load data it is these four thingsonly right so if I go ahead and assignthis over here right so if you want toread let's say housing data you can readthat rightnow if you want to go ahead and if youwant to explore this data you can alsonow this will also be mostly image dataonly so

(Editor: go)