deepness.processing.models.detector.Detector¶

class Detector(model_file_path: str)¶

Bases: ModelBase

Class implements object detection features

Detector model is used for detection of objects in images. It is based on YOLOv5/YOLOv7 models style.

Methods

`check_loaded_model_outputs`	Check if model outputs are valid. Valid model are: - has 1 or 2 outputs layer - output layer shape length is 3 - batch size is 1.
`crop_mask`
`get_channel_name`	Get channel name by id if exists in model metadata
`get_class_display_name`	Get class display name
`get_detector_type`	Get detector type from metadata if exists
`get_input_shape`	Get shape of the input for the model
`get_input_size_in_pixels`	Get number of input pixels in x and y direction (the same value)
`get_metadata_detection_confidence`	Get detection confidence from metadata if exists
`get_metadata_detection_iou_threshold`	Get detection iou threshold from metadata if exists
`get_metadata_model_type`	Get model type from metadata
`get_metadata_regression_output_scaling`	Get regression output scaling from metadata if exists
`get_metadata_resolution`	Get resolution from metadata if exists
`get_metadata_segmentation_small_segment`	Get segmentation small segment from metadata if exists
`get_metadata_segmentation_threshold`	Get segmentation threshold from metadata if exists
`get_metadata_standarization_parameters`	Get standardization parameters from metadata if exists
`get_metadata_tile_size`	Get tile size from metadata if exists
`get_metadata_tiles_overlap`	Get tiles overlap from metadata if exists
`get_model_batch_size`	Get batch size of the model
`get_model_type_from_metadata`	Get model type from metadata
`get_number_of_channels`	Returns number of channels in the input layer
`get_number_of_output_channels`	Get number of output channels
`get_output_shapes`	Get shapes of the outputs for the model
`get_outputs_channel_names`	Get class names from metadata
`non_max_suppression_fast`	Apply non-maximum suppression to bounding boxes
`postprocessing`	Postprocess model output
`preprocessing`	Preprocess the batch of images for the model (resize, normalization, etc)
`process`	Process a single tile image
`process_mask`
`set_inference_params`	Set inference parameters
`set_model_type_param`	Set model type parameters
`sigmoid`
`xywh2xyxy`	Convert bounding box from (x,y,w,h) to (x1,y1,x2,y2) format

Attributes

`confidence`	Confidence threshold
`iou_threshold`	IoU threshold
`model_type`	Model type

check_loaded_model_outputs()¶

Check if model outputs are valid. Valid model are:

has 1 or 2 outputs layer

output layer shape length is 3

batch size is 1

confidence¶

Confidence threshold

Type:: float

get_channel_name(layer_id: int, channel_id: int) → str¶

Get channel name by id if exists in model metadata

Parameters:: channel_id (int) – Channel id (means index in the output tensor)
Returns:: Channel name or empty string if not found
Return type:: str

classmethod get_class_display_name()¶

Get class display name

Returns:: Class display name
Return type:: str

get_detector_type() → str | None¶

Get detector type from metadata if exists

Returns string value of DetectorType enum or None if not found¶

Optional[str]: Detector type or None if not found

get_input_shape() → tuple¶

Get shape of the input for the model

Returns:: Shape of the input (batch_size, channels, height, width)
Return type:: tuple

get_input_size_in_pixels() → int¶

Get number of input pixels in x and y direction (the same value)

Returns:: Number of pixels in x and y direction
Return type:: int

get_metadata_detection_confidence() → float | None¶

Get detection confidence from metadata if exists

Returns:: Detection confidence or None if not found
Return type:: Optional[float]

get_metadata_detection_iou_threshold() → float | None¶

Get detection iou threshold from metadata if exists

Returns:: Detection iou threshold or None if not found
Return type:: Optional[float]

get_metadata_model_type() → str | None¶

Get model type from metadata

Returns:: Model type or None if not found
Return type:: Optional[str]

get_metadata_regression_output_scaling() → float | None¶

Get regression output scaling from metadata if exists

Returns:: Regression output scaling or None if not found
Return type:: Optional[float]

get_metadata_resolution() → float | None¶

Get resolution from metadata if exists

Returns:: Resolution or None if not found
Return type:: Optional[float]

get_metadata_segmentation_small_segment() → int | None¶

Get segmentation small segment from metadata if exists

Returns:: Segmentation small segment or None if not found
Return type:: Optional[int]

get_metadata_segmentation_threshold() → float | None¶

Get segmentation threshold from metadata if exists

Returns:: Segmentation threshold or None if not found
Return type:: Optional[float]

get_metadata_standarization_parameters() → StandardizationParameters | None¶

Get standardization parameters from metadata if exists

Returns:: Standardization parameters or None if not found
Return type:: Optional[StandardizationParameters]

get_metadata_tile_size() → int | None¶

Get tile size from metadata if exists

Returns:: Tile size or None if not found
Return type:: Optional[int]

get_metadata_tiles_overlap() → int | None¶

Get tiles overlap from metadata if exists

Returns:: Tiles overlap or None if not found
Return type:: Optional[int]

get_model_batch_size() → int | None¶

Get batch size of the model

Returns:: Batch size or None if not found (dynamic batch size)
Return type:: Optional[int] | None

classmethod get_model_type_from_metadata(model_file_path: str) → str | None¶

Get model type from metadata

Parameters:: model_file_path (str) – Path to the model file
Returns:: Model type or None if not found
Return type:: Optional[str]

get_number_of_channels() → int¶

Returns number of channels in the input layer

Returns:: Number of channels in the input layer
Return type:: int

get_number_of_output_channels()¶

Get number of output channels

Returns:: Number of output channels
Return type:: int

get_output_shapes() → List[tuple]¶

Get shapes of the outputs for the model

Returns:: Shapes of the outputs (batch_size, channels, height, width)
Return type:: List[tuple]

get_outputs_channel_names() → List[List[str]] | None¶

Get class names from metadata

Returns:: List of class names for each model output or None if not found
Return type:: List[List[str]] | None

iou_threshold¶

IoU threshold

Type:: float

model_type: DetectorType | None¶

Model type

Type:: DetectorType

static non_max_suppression_fast(boxes: ndarray, probs: ndarray, iou_threshold: float) → List¶

Apply non-maximum suppression to bounding boxes

Based on: https://github.com/amusi/Non-Maximum-Suppression/blob/master/nms.py

Parameters:

boxes (np.ndarray) – Bounding boxes in (x1,y1,x2,y2) format
probs (np.ndarray) – Confidence scores
iou_threshold (float) – IoU threshold

Returns:

List of indexes of bounding boxes to keep

Return type:

List

postprocessing(model_output)¶

Postprocess model output

NOTE: Maybe refactor this, as it has many added layers of checks which can be simplified.

Parameters:: model_output (list) – Model output
Returns:: Batch of lists of detections
Return type:: list

preprocessing(tiles_batched: ndarray) → ndarray¶

Preprocess the batch of images for the model (resize, normalization, etc)

Parameters:: image (np.ndarray) – Batch of images to preprocess (N,H,W,C), RGB, 0-255
Returns:: Preprocessed batch of image (N,C,H,W), RGB, 0-1
Return type:: np.ndarray

process(tiles_batched: ndarray)¶

Process a single tile image

Parameters:: img (np.ndarray) – Image to process ([TILE_SIZE x TILE_SIZE x channels], type uint8, values 0 to 255)
Returns:: Single prediction
Return type:: np.ndarray

set_inference_params(confidence: float, iou_threshold: float)¶

Set inference parameters

Parameters:

confidence (float) – Confidence threshold
iou_threshold (float) – IoU threshold

set_model_type_param(model_type: DetectorType)¶

Set model type parameters

Parameters:: model_type (str) – Model type

static xywh2xyxy(x: ndarray) → ndarray¶

Convert bounding box from (x,y,w,h) to (x1,y1,x2,y2) format

Parameters:: x (np.ndarray) – Bounding box in (x,y,w,h) format
Returns:: Bounding box in (x1,y1,x2,y2) format
Return type:: np.ndarray