RNN

RNN Notes ( in development )

Recurrent Neural Network

• Input
• Hidden State
• Output

We can process a sequence of vectors x by applying a recurrence of formula at every time step.

ht = fw(ht-1, xt)

Example of rnn for captcha identification

first let's define a dataset, we'll use the dataset base class from pytorch. The dataloader is being used to stack the images together into batches. we will use a collate_fn method to fix the issue of variable length

here is a sample image :

defining the dataset class

import os



class CaptchaDataset(Dataset):
    def __init__(self) -> None:
        self.data_dir = "" # make sure data_dir ends with forward slash
        self.filenames = [self.data_dir + item for item in os.listdir(self.data_dir)]
        return 


    def __len__(self) -> int:
        return len(self.filenames)


    def __getitem__(self, idx):
        """
        load image
        convert to single channel 
        return img in tensor format, label
        """

        return