imlab.extractor.Extractor

class imlab.extractor.Extractor(weight: ~typing.Optional[str] = None, model_type: str = 'yoloV7', image_size: int = 640, model: ~typing.Optional[object] = None, preprocess: callable = <function preprocess>, decode: callable = <function decode>, classes: list = [])

Bases: Wrapper

Extractor. Class to detect entities from picture

__init__(weight: ~typing.Optional[str] = None, model_type: str = 'yoloV7', image_size: int = 640, model: ~typing.Optional[object] = None, preprocess: callable = <function preprocess>, decode: callable = <function decode>, classes: list = []) → None

__init__.

Parameters

• weight (str) –

• model_type (str) –

• image_size (int) –

• model (object) –

• preprocess (callable) –

• decode (callable) –

• classes (list) –

Return type

None

Methods

__init__([weight, model_type, image_size, ...])	__init__.
dump(file_handler)	dump.
get_ratio(img1, img2)	get_ratio.
norm_input(image)	norm_input.
open_img(image)	open_img.
predict(images[, num])	predict.
rescale(boxes, gain, pad, img_size)	rescale.

dump(file_handler: object) → None

dump. dump the model

Parameters

file_handler (object) –

Return type

None

get_ratio(img1: (<class 'int'>, <class 'int'>), img2: (<class 'int'>, <class 'int'>)) -> (<class 'float'>, (<class 'float'>, <class 'float'>))

get_ratio.

Parameters

• img1 ((int, int)) –

• img2 ((int, int)) –

Return type

(float, (float, float))

norm_input(image: object) → FloatTensor

norm_input.

Parameters

image (object) –

Return type

torch.FloatTensor

open_img(image: object) → ndarray

open_img.

Parameters

image (object) –

Return type

np.ndarray

predict(images: object | list, num: int = 1) → list[tuple]

predict. find entities from entities

Parameters

• images (object | list) –

• num (int) – number of result per bouding box

Return type

list[tuple]

rescale(boxes: [<class 'torch.FloatTensor'>], gain: float, pad: (<class 'float'>, <class 'float'>), img_size: (<class 'int'>, <class 'int'>)) → [<class 'torch.FloatTensor'>]

rescale.

Parameters

• boxes ([torch.FloatTensor]) –

• gain (float) –

• pad ((float, float)) –

• img_size ((int, int)) –

Return type

[torch.FloatTensor]