The basic model which is used everywhere and have many usecases in computer vision. Happy Reading 🙂

VGG-Net(16)

You will find this net in every text book. But, they will just talk about architecture and some training hyperparameter. But, devil is in details. If we are not seeing its implementation then we haven’t gotten every bit of information. I don’t have time to implement it, and i guess you also don’t have that much time so, let’s see some other’s work.

VGG-Net(16) implementation

By just seeing this we don’t get information like, was there batchnormlization. Dropout layer is used or not. In machine learning, if we want to learn, then from my prespective we need to see other’s model in detail. In detail means you should know what learning rate they have used. Just architecture is not enough.

Now we know VGG that using 16 layer of convolution and regularization technique we can classify an object in 100 or 1000 different classes. Now, when we see other applications like crop or fruit breed classification. Or skin diseases idenfication etc. We can start with VGG and improve it by adding something here and there.

Nobody is going to read this. This is for myself. So, i can revise that what i have learned.