i would start with letter reconition, because for words you'll need a huge dataset, since the size of the unique words set is about 170,000 , for each one of them you need thousands of samples and labels. but for letters the dataset can be just for the 26 letters in the english alphabet.
enjoy, and support us by liking the video and subscribing to the channel.
Could you please upload the part 2 of Lane detection using a deep learning model?🥺
Can we use it for words recognition?
i would start with letter reconition, because for words you'll need a huge dataset, since the size of the unique words set is about 170,000 , for each one of them you need thousands of samples and labels.
but for letters the dataset can be just for the 26 letters in the english alphabet.
Can we use it for alphabet recognition?
i think you can use this model's architecture for spoken letter recongnition as well yes.