decrypt

decrypt

decrypt: predict labels for captcha images using a trained CRNN model.

Functions

Name	Description
decrypt	Predict labels for one or more captcha images.

decrypt.decrypt(
    files,
    model=None,
    mask=None,
    case_sensitive=True,
    length=None,
    device=None,
)

Predict labels for one or more captcha images.

Name	Type	Description	Default
files	Union[PathLike, Sequence[PathLike], Captcha]		required
model	Optional[Union[CRNN, str]]	or `None`. When `None`, the default unified checkpoint (:data:`txtcaptcha.hub.DEFAULT_REPO_ID`) is downloaded from the HF Hub on first call and cached thereafter — the cache lives in `~/.cache/huggingface/hub` and is shared with every other `huggingface_hub` user on the machine.	`None`
mask	MaskType	characters (e.g. `['a','b','c']`), a regex character class (e.g. `'[0-9a-z]'` or `'0-9'`), or a compiled `re.Pattern`.	`None`
case_sensitive	bool	are accepted; output is also lowercased.	`True`
length	Optional[int]	characters. Uses an exact CTC DP over constrained paths — strictly better than greedy when you know the captcha length up-front, since it never emits a wrong-length prediction. Leave as `None` to use the (variable-length) greedy CTC decoder.	`None`
device	Optional[Union[str, torch.device]]		`None`