DetectorDNNOpenCV¶

class DetectorDNNOpenCV(self)¶

Bases: pybind11_object

This class is a wrapper over the OpenCV DNN module and specialized to handle object detection task.

This class supports the following networks dedicated to object detection:

Faster-RCNN, see usage to detect objects belonging to the COCO dataset using dnn_supported_faster_rcnn network
SSD MobileNet, see usage to detect objects belonging to the COCO dataset using dnn_supported_mobilenet_ssd network
ResNet 10, see usage for dnn_usecase_face_detection
Yolo v3, see usage to detect objects belonging to the COCO dataset using dnn_supported_yolov3 network
Yolo v4, see usage to detect objects belonging to the COCO dataset using dnn_supported_yolov4 network
Yolo v5, see usage to detect objects belonging to the COCO dataset using dnn_supported_yolov5 network
Yolo v7, see usage to detect objects belonging to the COCO dataset using dnn_supported_yolov7 network
Yolo v8, see usage to detect objects belonging to the COCO dataset using dnn_supported_yolov8 network

This class can be initialized from a JSON file if ViSP has been compiled with NLOHMANN JSON (see soft_tool_json to see how to do it). Examples of such JSON files can be found in the tutorial folder.

Note

See tutorial-detection-dnn

Methods

`__init__`
`detect`	Overloaded function.
`dnnResultsParsingTypeFromString`	Cast a name of class of parsing method into a vpDetectorDNNOpenCV::DNNResultsParsingType .
`dnnResultsParsingTypeToString`
`getAvailableDnnResultsParsingTypes`	Get the list of the parsing methods / types of DNNs supported by the vpDetectorDNNOpenCV class.
`getNetConfig`
`initFromJSON`	param jsonPath:
`parseClassNamesFile`	Parse the designated file that contains the list of the classes the network can detect.
`readNet`	Read a network, see OpenCV readNet documentation for more information.
`saveConfigurationInJSON`	Save the network configuration in a JSON file.
`setConfidenceThreshold`	Set confidence threshold to filter the detections.
`setDetectionFilterSizeRatio`	Set the size ratio used in the filterDetection method.
`setInputSize`	Set dimension to resize the image to the input blob.
`setMean`	Set mean subtraction values.
`setNMSThreshold`	Set Non-Maximum Suppression threshold, used to filter multiple detections at approximatively the same location.
`setNetConfig`
`setPreferableBackend`	Set preferable backend for inference computation.
`setPreferableTarget`	Set preferable target for inference computation.
`setScaleFactor`	Set scale factor to normalize the range of pixel values.
`setSwapRB`	If true, swap R and B channel for mean subtraction.

Inherited Methods

Operators

`__doc__`
`__init__`
`__module__`
`__repr__`

Attributes

`COUNT`
`FASTER_RCNN`
`RESNET_10`
`SSD_MOBILENET`
`USER_SPECIFIED`
`YOLO_V11`
`YOLO_V3`
`YOLO_V4`
`YOLO_V5`
`YOLO_V7`
`YOLO_V8`
`__annotations__`

class DNNResultsParsingType(self, value: int)¶

Bases: pybind11_object

Enumeration listing the types of DNN for which the ** vpDetectorDNNOpenCV ** furnishes the methods permitting to parse the raw detection data into ** vpDetectorDNNOpenCV::DetectedFeatures2D ** .

Values:

USER_SPECIFIED: The user will give a pointer towards the parsing method to use to parse the raw data resulting from the detection step.
FASTER_RCNN: The ** vpDetectorDNNOpenCV ** object will use the parsing method corresponding to a Faster-RCNN DNN. See ** vpDetectorDNNOpenCV::postProcess_FasterRCNN ** for more information.
SSD_MOBILENET: The ** vpDetectorDNNOpenCV ** object will use the parsing method corresponding to a SSD MobileNet DNN. See ** vpDetectorDNNOpenCV::postProcess_SSD_MobileNet ** for more information.
RESNET_10: The ** vpDetectorDNNOpenCV ** object will use the parsing method corresponding to a ResNet 10 DNN. See ** vpDetectorDNNOpenCV::postProcess_ResNet_10 ** for more information.
YOLO_V3: The ** vpDetectorDNNOpenCV ** object will use the parsing method corresponding to a YoloV3 DNN. See ** vpDetectorDNNOpenCV::postProcess_YoloV3_V4 ** for more information.
YOLO_V4: The ** vpDetectorDNNOpenCV ** object will use the parsing method corresponding to a YoloV4 DNN. See ** vpDetectorDNNOpenCV::postProcess_YoloV3_V4 ** for more information.
YOLO_V5: The ** vpDetectorDNNOpenCV ** object will use the parsing method corresponding to a YoloV5 DNN. See ** vpDetectorDNNOpenCV::postProcess_YoloV5_V7 ** for more information.
YOLO_V7: The ** vpDetectorDNNOpenCV ** object will use the parsing method corresponding to a YoloV7 DNN. See ** vpDetectorDNNOpenCV::postProcess_YoloV5_V7 ** for more information.
YOLO_V8: The ** vpDetectorDNNOpenCV ** object will use the parsing method corresponding to a YoloV8 DNN. See ** vpDetectorDNNOpenCV::postProcess_YoloV8 ** for more information.
COUNT: The number of parsing method that come along with the ** vpDetectorDNNOpenCV ** class.

__and__(self, other: object) → object¶

__eq__(self, other: object) → bool¶

__ge__(self, other: object) → bool¶

__getstate__(self) → int¶

__gt__(self, other: object) → bool¶

__hash__(self) → int¶

__index__(self) → int¶

__init__(self, value: int)¶

__int__(self) → int¶

__invert__(self) → object¶

__le__(self, other: object) → bool¶

__lt__(self, other: object) → bool¶

__ne__(self, other: object) → bool¶

__or__(self, other: object) → object¶

__rand__(self, other: object) → object¶

__ror__(self, other: object) → object¶

__rxor__(self, other: object) → object¶

__setstate__(self, state: int) → None¶

__xor__(self, other: object) → object¶

property name : str¶

__init__(self)¶

detect(*args, **kwargs)¶

Overloaded function.

detect(self: visp._visp.detection.DetectorDNNOpenCV, I: visp._visp.core.ImageGray, output: list[visp._visp.detection.DetectorDNNOpenCV.DetectedFeatures2D]) -> tuple[bool, list[visp._visp.detection.DetectorDNNOpenCV.DetectedFeatures2D]]

Object detection using OpenCV DNN module.

Parameters:

I: Input image.
output: Vector of detections, whichever class they belong to.

Returns:

A tuple containing:

false if there is no detection, true otherwise.
output: Vector of detections, whichever class they belong to.

detect(self: visp._visp.detection.DetectorDNNOpenCV, I: visp._visp.core.ImageGray, output: dict[str, list[visp._visp.detection.DetectorDNNOpenCV.DetectedFeatures2D]]) -> bool

Parameters:

I: Input image.
output: map where the name of the class is used as key and whose value is a vector of detected 2D features that belong to the class.

Returns:

false if there is no detection.

detect(self: visp._visp.detection.DetectorDNNOpenCV, I: visp._visp.core.ImageGray, output: list[tuple[str, list[visp._visp.detection.DetectorDNNOpenCV.DetectedFeatures2D]]]) -> tuple[bool, list[tuple[str, list[visp._visp.detection.DetectorDNNOpenCV.DetectedFeatures2D]]]]

Parameters:

I: Input image.
output: vector of pairs <name_of_the_class, vector_of_detections>

Returns:

A tuple containing:

false if there is no detection.
output: vector of pairs <name_of_the_class, vector_of_detections>

detect(self: visp._visp.detection.DetectorDNNOpenCV, I: visp._visp.core.ImageRGBa, output: list[visp._visp.detection.DetectorDNNOpenCV.DetectedFeatures2D]) -> tuple[bool, list[visp._visp.detection.DetectorDNNOpenCV.DetectedFeatures2D]]

Object detection using OpenCV DNN module.

Parameters:

I: Input image.
output: Vector of detections, whichever class they belong to.

Returns:

A tuple containing:

false if there is no detection, true otherwise.
output: Vector of detections, whichever class they belong to.

detect(self: visp._visp.detection.DetectorDNNOpenCV, I: visp._visp.core.ImageRGBa, output: dict[str, list[visp._visp.detection.DetectorDNNOpenCV.DetectedFeatures2D]]) -> bool

Object detection using OpenCV DNN module.

Parameters:

I: Input image.
output: map where the name of the class is used as key and whose value is a vector of detected 2D features that belong to the class.

Returns:

false if there is no detection.

detect(self: visp._visp.detection.DetectorDNNOpenCV, I: visp._visp.core.ImageRGBa, output: list[tuple[str, list[visp._visp.detection.DetectorDNNOpenCV.DetectedFeatures2D]]]) -> tuple[bool, list[tuple[str, list[visp._visp.detection.DetectorDNNOpenCV.DetectedFeatures2D]]]]

Object detection using OpenCV DNN module.

Parameters:

I: Input image.
output: vector of pairs <name_of_the_class, vector_of_detections>

Returns:

A tuple containing:

false if there is no detection, true otherwise.
output: vector of pairs <name_of_the_class, vector_of_detections>

detect(self: visp._visp.detection.DetectorDNNOpenCV, I: cv::Mat, output: list[visp._visp.detection.DetectorDNNOpenCV.DetectedFeatures2D]) -> tuple[bool, list[visp._visp.detection.DetectorDNNOpenCV.DetectedFeatures2D]]

Object detection using OpenCV DNN module.

Parameters:

I: Input image.
output: Vector of detections, whichever class they belong to.

Returns:

A tuple containing:

false if there is no detection.
output: Vector of detections, whichever class they belong to.

detect(self: visp._visp.detection.DetectorDNNOpenCV, I: cv::Mat, output: dict[str, list[visp._visp.detection.DetectorDNNOpenCV.DetectedFeatures2D]]) -> bool

Object detection using OpenCV DNN module.

Parameters:

I: Input image.
output: map where the name of the class is used as key and whose value is a vector of detected 2D features that belong to the class.

Returns:

false if there is no detection.

detect(self: visp._visp.detection.DetectorDNNOpenCV, I: cv::Mat, output: list[tuple[str, list[visp._visp.detection.DetectorDNNOpenCV.DetectedFeatures2D]]]) -> tuple[bool, list[tuple[str, list[visp._visp.detection.DetectorDNNOpenCV.DetectedFeatures2D]]]]

Object detection using OpenCV DNN module.

Parameters:

I: Input image.
output: vector of pairs <name_of_the_class, vector_of_detections>

Returns:

A tuple containing:

false if there is no detection.
output: vector of pairs <name_of_the_class, vector_of_detections>

static dnnResultsParsingTypeFromString(name: str) → visp._visp.detection.DetectorDNNOpenCV.DNNResultsParsingType¶

Cast a name of class of parsing method into a ** vpDetectorDNNOpenCV::DNNResultsParsingType ** . The naming convention is the following:

only lowercases

static dnnResultsParsingTypeToString(type: visp._visp.detection.DetectorDNNOpenCV.DNNResultsParsingType) → str¶

static getAvailableDnnResultsParsingTypes() → str¶

Get the list of the parsing methods / types of DNNs supported by the ** vpDetectorDNNOpenCV ** class.

Returns:: std::string The list of the supported parsing methods / types of DNNs.

getNetConfig(self) → visp._visp.detection.DetectorDNNOpenCV.NetConfig¶

initFromJSON(self, jsonPath: str) → None¶

Parameters:

jsonPath: str¶

static parseClassNamesFile(filename: str) → list[str]¶

Parse the designated file that contains the list of the classes the network can detect. The class names must either be indicated in an array of string in YAML format, or one name by row (without quotes) as described in NetConfig::parseClassNamesFile() .

Note

See NetConfig::parseClassNamesFile() .

Parameters:

filename: str¶: The path towards the file.

Returns:

std::vector<std::string> The list of class names.

readNet(self: visp._visp.detection.DetectorDNNOpenCV, model: str, config: str =, framework: str =) → None¶

Read a network, see OpenCV readNet documentation for more information.

Parameters:

model

Path to a binary file of model containing trained weights. The following file extensions are expected for models from different frameworks:

*.caffemodel (Caffe, http://caffe.berkeleyvision.org/ )
*.pb (TensorFlow, https://www.tensorflow.org/ )
*.t7 | *.net (Torch, http://torch.ch/ )
*.weights (Darknet, https://pjreddie.com/darknet/ )
*.bin (DLDT, https://software.intel.com/openvino-toolkit )
*.onnx (ONNX, https://onnx.ai/ )

config

Path to a text file of model containing network configuration. It could be a file with the following extensions:

*.prototxt (Caffe, http://caffe.berkeleyvision.org/ )
*.pbtxt (TensorFlow, https://www.tensorflow.org/ )
*.cfg (Darknet, https://pjreddie.com/darknet/ )
*.xml (DLDT, https://software.intel.com/openvino-toolkit )

framework

Optional name of an origin framework of the model. Automatically detected if it is not set.

saveConfigurationInJSON(self, jsonPath: str) → None¶

Save the network configuration in a JSON file.

Parameters:

jsonPath: str¶: Path towards the output JSON file .

setConfidenceThreshold(self, confThreshold: float) → None¶

Set confidence threshold to filter the detections.

Parameters:

confThreshold: float¶: Confidence threshold between [0, 1]

setDetectionFilterSizeRatio(self, sizeRatio: float) → None¶

Set the size ratio used in the filterDetection method. If <= 0., filterDetection is not used. The detections of class c for which the bbox area does not belong to [mean_class_c(Area) * sizeRatio ; mean_class_c(Area) / sizeRatio ] are removed from the list of detections.

Parameters:

sizeRatio: float¶: the size ratio used in the filterDetection method. If <= 0., filterDetection is not used.

setInputSize(self, width: int, height: int) → None¶

Set dimension to resize the image to the input blob.

Parameters:

width: int¶: If <= 0, blob width is set to image width
height: int¶: If <= 0, blob height is set to image height

setMean(self, meanR: float, meanG: float, meanB: float) → None¶

Set mean subtraction values.

Parameters:

meanR: float¶: Mean value for R-channel
meanG: float¶: Mean value for G-channel
meanB: float¶: Mean value for R-channel

setNMSThreshold(self, nmsThreshold: float) → None¶

Set Non-Maximum Suppression threshold, used to filter multiple detections at approximatively the same location.

Parameters:

nmsThreshold: float¶: Non-Maximum Suppression threshold between [0, 1]

setNetConfig(self, config: visp._visp.detection.DetectorDNNOpenCV.NetConfig) → None¶

setPreferableBackend(self, backendId: int) → None¶

Set preferable backend for inference computation. See OpenCV setPreferableBackend documentation for more information.

Parameters:

backendId: int¶: Backend identifier

setPreferableTarget(self, targetId: int) → None¶

Set preferable target for inference computation. See OpenCV setPreferableTarget documentation for more information.

Parameters:

targetId: int¶: Target identifier

setScaleFactor(self, scaleFactor: float) → None¶: Set scale factor to normalize the range of pixel values.

setSwapRB(self, swapRB: bool) → None¶: If true, swap R and B channel for mean subtraction. For instance when the network has been trained on RGB image format (OpenCV uses BGR convention).