In the lecture End to End deep learning, we had doubt in the below statement. What is process should we apply to compare two input images that feed to the Neural network ? How it can done ? Should we regression ? can u please help to explain about a bit? I dont why neural network needs to compare two images and not sure how it can be done ? As far programming assignment we dont see comparing two images fed to the network
Neural network does is it takes this input two images and it tells you if these two are the same person or not.
The short “high level” version of the answer is that you have to train a network to achieve whatever the goal is that you have. How that works will be determined by the cost or loss function that you choose. How face recognition algorithms work will be covered in detail in Course 4 Week 4. The basic idea is that you define a loss function that is designed to measure the “distance” between two faces. You then train your model on a labelled image dataset as usual to minimize the distance in the case that the pictures are of the same person.
Here’s a screenshot showing the first few lectures in Week 4 of Course 4. If you want to know more, you can either wait for Course 4 or watch a few of the lectures to get a preview.
