Captcha_Decoder

We start from taking a dataset of 1000 captcha images (200*50) from kaggle.[1] Each image has a 5 letter word that also contains numbers. The images have noise applied to them (blur and a line). The first thing that we’ll have to do after this is to convert these images to grayscale to get numeric data on which we can work on. Then we move to the cleaning part, since the images are blurred and with a line passing through the text, we’ll have to work around that and segment individual characters from the image and extract features required for training a Clustering algorithm (K-Means) which will be used to decode the captcha values.