techietrader

/

captcha_ocr

like 11

Follow

Techie Traders 6

yeah definitely,
_decode_cls_dims = {
"63": 63, "37_uppercase": 37,
"11": 11, "37_lowercase": 37
}

len_dim_pair = {
"v1": (6, "63"), "v2": (6, "11"), "v3": (5, "63"),
"v4": (6, "37_lowercase"), "v5": (5, "37_uppercase"),
"v6": (4, "63"), "v7": (6, "11"), "v8": (5, "37_uppercase")
}

char_sets = {
"63": "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789$",
"37_uppercase": "ABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789$",
"37_lowercase": "abcdefghijklmnopqrstuvwxyz0123456789$",
"11": "0123456789$"
}

Thank you. I have another question. If the number of characters in the captcha is different, is there a way to identify them?

yes there are ways but you need to do that in early phase of model training.
encode through cnn and decode by rnn being most traditional ones.