Hello everyone. In this video, we're going to get an overview on what our image search model is actually doing. We're going to take a look at the architecture and define broad brushes, what we're going to build. on the server side, as you can see in this image, we start with a training set, which is images and want to put them through a classical cnn classifier. Not nothing special about this. It's a very classic classification task, a neural net, and we put the image is in and we saved the image vectors as outputs.

Now, after training, the network, network layers themselves are going to be a feature representation of each image. We're going to take these high level features and save them into database. This database is going to store image vector pairs which we're going to Use and compare to user input images for search purposes. From the database, you can see that there's an arrow pointing down to the comparing training set vectors with output image vector image. And this is where the database images and user uploaded images converge. Let's take a look at how the uploaded images make their way here.

The user is going to upload an image onto a website, which we're going to create. And the image is going to go through a pre trained model as well through a CNN classifier is going to turn the image into an image vector and the same format as the ones that we're trained on. Then it's going to use one of two distance metrics that we have decided to use, which are going to be Hamming or cosine distance. Depending on those going to give us results, sorting other images in database compared to the uploaded one by how close they are in this vector space. So if we upload an image of a cat, it's going to look at images, which in this high level representation in a neural net, resemble cats, and hopefully going to be hearing useful results, meaning that if we upload an image of a cat, we get results of pictures of a lot of different cats.

Oh, is there any questions about this about how the architecture is supposed to work? You can leave them in the comments section and Otherwise, I'll see you in the next video.