Low Code/No Code AI Platform Teachable Machine: Testing the Model

Key Takeaways

foreign welcome to AI anytime channel so in the last video we used uh low code no code AI platform teachable machine to build a classifier okay we build a classification model with the help of teachable machine where we downloaded first uh brain MRI images from kaggle and then we uploaded those images the data set on teachable machine and we use the platform to train a model and then we exported that model you know locally right so you can see on my screen this this was the platform that we used and we also tested it out with ends with a sample image right so you can export the model from here that we have already exported okay so we'll so in this video we'll test it out we'll test that the python code that was given here if you see here let me come down okay the code snippet that we we can see uh which has been this has been given as an instruction on that how we can use that model okay if you really want to test it out locally so let me go back to folder I will create a new folder here I'll call it app because later we will also be build a streamlit application you know so in this I'll first create I'll copy this test dot PNG file that we need put it inside app and in app I will open this in terminal good and I will create a virtual environment so Python 3 iPhone name VNV and Dot VNV it will create a virtual environment for me I will go inside VNV and Bin if you if you are on Windows you might be using power cell or terminal you have to go to scripts not in bing and then uh Source activate to activate this virtual environment you can see in the right hand side we have app now which has been activated okay then I will go to the root directory the main directory that we have right and then I'll create uh maybe let's go to vs code and create the file there okay uh well first let me turn off all these notifications that I am getting so you can see I am in this app folder or the directory and then I have a VNV there and I have a sample file that we can test we first need to create a requirements file requirements.txt okay because we have to use a few libraries okay there are a few dependencies that are needed to run that code snippet that we have you know we have seen on the the teachable machine platform Okay so they they have given the Keras model but now Keras comes uh by by default when you install tensorflow what I'll do I will do tensorflow CPU I'm not going to use GPU for this inference so I will write tensorflow hyphen CPU to just download the CPU version of tensorflow very minimal modules required right to run that image classifier if you if you are on GP GPU you can have tensorflow that and though I have GPU but I will not use it okay so I'll just use tensorflow CPU and then uh what else we will need uh numpy and below right these are the things which required let me go back to this code and see yeah so what we will do now we'll just uh first copy this code from here and I'll get back to vs code and meanwhile let me first do a pip install so P3 install hyphen R requirements.txt we have to first install all the dependencies right so it will take little time because tensorflow CPU you can see right it says around 225 MB inside right anyway uh what next so we'll create a test.py file where we will test this python code and see if he's running as it is okay or we have to make some changes in the code in we will also create a streamlit application later on and then we'll continue it and then we'll deploy it but um first first thing that I can see that we need this couple of files in the directory okay the Keras model or levels txt okay then let me first see this has been downloaded then let me go back if you if I go back you can see this converted Keras folder that I have I'll copy this to come inside app and I'll paste it so now you can see I have this Keras underscore model.h5 and the labels.txt if you click on levels.txt we have 0 which means no and one means which is yes the binary classification problem where there are two classes in Kira's model I can see this has to be renamed okay underscore H5 levels txt install below instead of pil this is a good well this is a good suggestion anyway so we have installed pillow if you see here and then we have numpy the model you can see the model has been loaded here with Keras uh modules load underscore model then your class name then we are you can see over here a very very well documented code okay they are determining the first position in the safety pose the Tuple and RGB and the sizes uh 224 across 224 the channel and then we are you have to give this image path so let me change the image path so it will be test.png so test dot PNG the sample file that we have and to convert the rdb if it's not and size 224 to 24 then we are turning the email into a numpy array so we can send that numpy array uh to the train model to perform the prediction step okay and we are normalizing the arrays you can see it over here with numpy DOT flow 32 and then then we are loading the email into the array and then running the inference that model that we have that we have you know imported over here and then we are getting this ARG max value the maximum of the classes that we have and then class name and prediction pretty much good so let's do one thing we have requirements of txt and everything so let's run this okay so what I will do I will run this okay perfect so when I ran the program or the code I will see it's a python file right so pretty much simple Python and the file name Dot py and you can see the class which says one yes it means tumor has been detected in this image that the test PNG file that we have you can see it over here and this also gives us a Confidence Code around 0.99 which mean 99 percent right have been detected so what we will do next now okay we will take this code and we'll create a streamlit application we will upload couple of more images and see when we are inferencing it okay so what we did we took the code snippet from teachable machine we had a sample image and then we created a requirement.txt file we installed all the dependencies or libraries and then we tested it out okay with the code snippet that we here in test.pi now in the next video we will create a streamlined application okay for uh integrating this model into a UI okay so if you like the content like the video please like and comment your thoughts or views of feedback in the comment box and please share this channel uh with your friends and in your peer thank you so much