Deployment of Deep Learning Model using Flask

Key Takeaways

hello all today we will be discussing how we can deploy a deep learning model with the help of flask in this particular example I'm basically going to use a transfer learning process which is called as resonant and I'm trying to deploy that I will deploy that using flask and before going head guys let me just show you a demo how it looks like resonant is basically the transfer learning the most of the technique that I basically used as resonate it is a transfer learning kind of technique where I basically loaded the weights of imagine a Tony now if you don't know about Imogen Imogen it is basically competition where you know different kind of neural networks with respect to like which is 16 we have we are like resonant we have like less intense named mobile in it so all these things they have come over here like they're basically want the competition in the image classification so immediate classification is basically about categorizing thousand different categories so let us just see an example suppose I go and choose a particular image for partners and if I do the prediction it will basically show me a giant underscore panda and similarly if I go and see for a bird so this looks like Macau so this Macau is nothing but it is a bird it's quite famous in the equatorial region you can see over here this is this was a demo thick little mesh but the front end of my behalf and I'm actually interacting with an API which is hosted using the flask and you can also choose different different images again let us see and just try to open up a snake image and try to predict this is like side window this is also a kind of snake Omni basically showing the name of the snake you can see that Sidewinder is also that and this is the demo all about and what we are going to do is that I'm just going to show you the code and how you can actually deploy this particular process reinvade guys there it is please to import all this code will be given in the github it is completely free for you or make sure you utilize it and try to practice by your own okay so initially you just import all these libraries which is which is with respect to chaos and all the other nice real IDs then what do you have is that just define the flask app this is simple because in my previous two videos I have also shown you how you can deploy a machine learning model and an NLP model so you should be pretty much clear if you have seen that you not seen that this is well and good when to see to it it just hardly five to ten minutes then what I do is that initially now initially my model is not created again I have to import that model and again when I am importing ResNet I will basically be getting this model from the corrals or application so what you can do is that just type down from Carol wrote application ResNet importer eyes net 50 and here you initialize the resonate 50 with the weights imaged that's it as soon as you get that value in your model your model will be created you just say model dot C and just give the path where you want to save the model yeah I've saved the model currently inside my model folders so here it is model resonate five model VG 16 if I have tried with both with ResNet and VG 16 it worked pretty well with both this algorithm scales so what you can do is that once you do this right you don't have to run it again and again because your model will get three bits so you can just comment down this particular code okay then you have to uncomment this particular code why you aren't commenting this particular code basically means that you are trying to read the h5 file directly from the model folder so here it is you are just reading it from the model folder then you do load more load model and model underscore predict function this was basically loading your model okay the next thing is that how do you define your approach this app route is basically for a root folder which will be like a get which is not it will which will which will do nothing but just show you their homepage which is present in index dot HTML index dot HTML is basically present inside this template folder or so template folder is basically having in this dot HTML so by default the render template will be there is not a stable for your hood but now the next thing is that for the predacons which is my API what I have to do is that what happen is that I just created a function for an upload as soon as the person uploads the picture in the web app what I do is that first of all I store that image inside a path which is called as uploads I store this image over here so you can see all the images get stored and that is the code this is the code which will store that particular image as soon as I stole this particular image you just have to go and hit on model and this will predict inside this I will just give the file path which is my image path and then my model which I have loaded earlier if I go and see what is my model and the scope predict yeah inside model and this could predict what I am doing is that unloading the image from the image path then I'm converting into an array finally I am expanding the dimensions and I am doing the Preferences input this is basically a normal step that we usually do and finally we do model dot credit so as we do the prediction we get the prediction over here and based on that we are just won't select the categories of the prediction and try to convert that back into string by using decoder underscore prediction this is a normal or ResNet and it is available in the Kalos documentation also and finally my main function from which I'm actually running Maya that's it this quite simple just hardly any lines of code in this particular file you just have all your folders set all you have to do is then run your front that's it in order to run it what you have to do is that just open up on the front again so hang up on the front here what you have to do is that just go to this particular part off in this path you know D Drive I'll say C D and here you just have to write fight them I'm dot T one I said soon as you press ENTER I'm not going to press ENTER because I've already my back end it is a warning you can see in my another command prompt it is already running in this particular you just have to press ENTER then you will suddenly get your local address and port number from where it is funny and after that you just have to open this you know just load it once and then you can choose whatever you want you see do the prediction straight - this is a kind of dog breed of dog basically so you can see it over here your images and this is your breed of dog I show one I had also uploaded one thing but let's see that there's a bad fighter yes in this board a lot so what am i J suffer I don't think so this is categorized properly but that is what this particular model says so this was about this particular session guys I hope you liked this particular video make sure you subscribe the channel like comment share with all your friends whoever required this I try to search and the internet a lot of things regarding how to acquire deep learning model I was not able to get a lot but as we go ahead and deployment part I'm also going to show you a CDC CIC deployment with respect to sure where and everything will be automated you're the coordinate gathering your model processing your feature engineering your feature selection everything every steps I find their deployment part make sure you subscribe the channel guys and I hope you are liking this particular video I'll see y'all in the next video have a great day ahead god bless you all thank you one and all