In Chapter 6, Visual Search Using Transfer Learning, we learned how to do a visual search on your local PC. The method uses passing an image through a neural network, such as VGG16 or ResNet, and converting it into an image vector by removing the last fully connected layer and then comparing the image with other images of known classes in the database to find the nearest neighbor match and then displaying the results.
We started with 200 images in our example, but if the number of images goes to 1 million and the results have to be accessed from a web page, storing the images locally will not make sense. In these cases, cloud storage is optimal. In that case, instead of storing the images in the cloud, we can store the image vector, and then when the user uploads an image, convert the image into a vector and send it to the cloud...
