Object Detection(目标检测)#

fastdeploy.vision.detection.PaddleDetPreprocessor#

class fastdeploy.vision.detection.PaddleDetPreprocessor(config_file)[source]#

Create a preprocessor for PaddleDetection Model from configuration file

Parameters

config_file – (str)Path of configuration file, e.g ppyoloe/infer_cfg.yml

disable_normalize()[source]#

This function will disable normalize in preprocessing step.

disable_permute()[source]#

This function will disable hwc2chw in preprocessing step.

run(input_ims)#

Process input image

Param

input_ims: (list of numpy.ndarray) The input images

Returns

list of FDTensor

use_cuda(enable_cv_cuda=False, gpu_id=-1)#

Use CUDA processors

Param

enable_cv_cuda: Ture: use CV-CUDA, False: use CUDA only

Param

gpu_id: GPU device id

fastdeploy.vision.detection.PaddleDetPostprocessor#

class fastdeploy.vision.detection.PaddleDetPostprocessor[source]#

Create a postprocessor for PaddleDetection Model

run(runtime_results)[source]#

Postprocess the runtime results for PaddleDetection Model

Param

runtime_results: (list of FDTensor)The output FDTensor results from runtime

Returns

list of ClassifyResult(If the runtime_results is predict by batched samples, the length of this list equals to the batch size)

set_nms_option(nms_option=None)[source]#

This function will enable decode and nms in postprocess step.

fastdeploy.vision.detection.PPYOLOE#

class fastdeploy.vision.detection.PPYOLOE(model_file, params_file, config_file, runtime_option=None, model_format=<ModelFormat.PADDLE: 1>)[source]#

Load a PPYOLOE model exported by PaddleDetection.

Parameters

  • model_file – (str)Path of model file, e.g ppyoloe/model.pdmodel

  • params_file – (str)Path of parameters file, e.g ppyoloe/model.pdiparams, if the model_fomat is ModelFormat.ONNX, this param will be ignored, can be set as empty string

  • config_file – (str)Path of configuration file for deployment, e.g ppyoloe/infer_cfg.yml

  • runtime_option – (fastdeploy.RuntimeOption)RuntimeOption for inference this model, if it’s None, will use the default backend on CPU

  • model_format – (fastdeploy.ModelForamt)Model format of the loaded model

batch_predict(images)[source]#

Detect a batch of input image list

Parameters

im – (list of numpy.ndarray) The input image list, each element is a 3-D array with layout HWC, BGR format

:return list of DetectionResult

clone()[source]#

Clone PPYOLOE object

Returns

a new PPYOLOE object

get_profile_time()#

Get profile time of Runtime after the profile process is done.

property postprocessor#

Get PaddleDetPostprocessor object of the loaded model

:return PaddleDetPostprocessor

predict(im)[source]#

Detect an input image

Parameters

im – (numpy.ndarray)The input image data, 3-D array with layout HWC, BGR format

Returns

DetectionResult

property preprocessor#

Get PaddleDetPreprocessor object of the loaded model

:return PaddleDetPreprocessor

fastdeploy.vision.detection.PPYOLO#

class fastdeploy.vision.detection.PPYOLO(model_file, params_file, config_file, runtime_option=None, model_format=<ModelFormat.PADDLE: 1>)[source]#

Load a PPYOLO model exported by PaddleDetection.

Parameters

  • model_file – (str)Path of model file, e.g ppyolo/model.pdmodel

  • params_file – (str)Path of parameters file, e.g ppyolo/model.pdiparams, if the model_fomat is ModelFormat.ONNX, this param will be ignored, can be set as empty string

  • runtime_option – (fastdeploy.RuntimeOption)RuntimeOption for inference this model, if it’s None, will use the default backend on CPU

  • model_format – (fastdeploy.ModelForamt)Model format of the loaded model

batch_predict(images)#

Detect a batch of input image list

Parameters

im – (list of numpy.ndarray) The input image list, each element is a 3-D array with layout HWC, BGR format

:return list of DetectionResult

clone()[source]#

Clone PPYOLO object

Returns

a new PPYOLO object

get_profile_time()#

Get profile time of Runtime after the profile process is done.

property postprocessor#

Get PaddleDetPostprocessor object of the loaded model

:return PaddleDetPostprocessor

predict(im)#

Detect an input image

Parameters

im – (numpy.ndarray)The input image data, 3-D array with layout HWC, BGR format

Returns

DetectionResult

property preprocessor#

Get PaddleDetPreprocessor object of the loaded model

:return PaddleDetPreprocessor

fastdeploy.vision.detection.PicoDet#

class fastdeploy.vision.detection.PicoDet(model_file, params_file, config_file, runtime_option=None, model_format=<ModelFormat.PADDLE: 1>)[source]#

Load a PicoDet model exported by PaddleDetection.

Parameters

  • model_file – (str)Path of model file, e.g picodet/model.pdmodel

  • params_file – (str)Path of parameters file, e.g picodet/model.pdiparams, if the model_fomat is ModelFormat.ONNX, this param will be ignored, can be set as empty string

  • config_file – (str)Path of configuration file for deployment, e.g ppyoloe/infer_cfg.yml

  • runtime_option – (fastdeploy.RuntimeOption)RuntimeOption for inference this model, if it’s None, will use the default backend on CPU

  • model_format – (fastdeploy.ModelForamt)Model format of the loaded model

batch_predict(images)#

Detect a batch of input image list

Parameters

im – (list of numpy.ndarray) The input image list, each element is a 3-D array with layout HWC, BGR format

:return list of DetectionResult

clone()[source]#

Clone PicoDet object

Returns

a new PicoDet object

get_profile_time()#

Get profile time of Runtime after the profile process is done.

property postprocessor#

Get PaddleDetPostprocessor object of the loaded model

:return PaddleDetPostprocessor

predict(im)#

Detect an input image

Parameters

im – (numpy.ndarray)The input image data, 3-D array with layout HWC, BGR format

Returns

DetectionResult

property preprocessor#

Get PaddleDetPreprocessor object of the loaded model

:return PaddleDetPreprocessor

fastdeploy.vision.detection.PaddleYOLOX#

class fastdeploy.vision.detection.PaddleYOLOX(model_file, params_file, config_file, runtime_option=None, model_format=<ModelFormat.PADDLE: 1>)[source]#

Load a YOLOX model exported by PaddleDetection.

Parameters

  • model_file – (str)Path of model file, e.g yolox/model.pdmodel

  • params_file – (str)Path of parameters file, e.g yolox/model.pdiparams, if the model_fomat is ModelFormat.ONNX, this param will be ignored, can be set as empty string

  • config_file – (str)Path of configuration file for deployment, e.g ppyoloe/infer_cfg.yml

  • runtime_option – (fastdeploy.RuntimeOption)RuntimeOption for inference this model, if it’s None, will use the default backend on CPU

  • model_format – (fastdeploy.ModelForamt)Model format of the loaded model

batch_predict(images)#

Detect a batch of input image list

Parameters

im – (list of numpy.ndarray) The input image list, each element is a 3-D array with layout HWC, BGR format

:return list of DetectionResult

clone()[source]#

Clone PaddleYOLOX object

Returns

a new PaddleYOLOX object

get_profile_time()#

Get profile time of Runtime after the profile process is done.

property postprocessor#

Get PaddleDetPostprocessor object of the loaded model

:return PaddleDetPostprocessor

predict(im)#

Detect an input image

Parameters

im – (numpy.ndarray)The input image data, 3-D array with layout HWC, BGR format

Returns

DetectionResult

property preprocessor#

Get PaddleDetPreprocessor object of the loaded model

:return PaddleDetPreprocessor

fastdeploy.vision.detection.YOLOv3#

class fastdeploy.vision.detection.PaddleYOLOX(model_file, params_file, config_file, runtime_option=None, model_format=<ModelFormat.PADDLE: 1>)[source]#

Load a YOLOX model exported by PaddleDetection.

Parameters

  • model_file – (str)Path of model file, e.g yolox/model.pdmodel

  • params_file – (str)Path of parameters file, e.g yolox/model.pdiparams, if the model_fomat is ModelFormat.ONNX, this param will be ignored, can be set as empty string

  • config_file – (str)Path of configuration file for deployment, e.g ppyoloe/infer_cfg.yml

  • runtime_option – (fastdeploy.RuntimeOption)RuntimeOption for inference this model, if it’s None, will use the default backend on CPU

  • model_format – (fastdeploy.ModelForamt)Model format of the loaded model

batch_predict(images)#

Detect a batch of input image list

Parameters

im – (list of numpy.ndarray) The input image list, each element is a 3-D array with layout HWC, BGR format

:return list of DetectionResult

clone()[source]#

Clone PaddleYOLOX object

Returns

a new PaddleYOLOX object

get_profile_time()#

Get profile time of Runtime after the profile process is done.

property postprocessor#

Get PaddleDetPostprocessor object of the loaded model

:return PaddleDetPostprocessor

predict(im)#

Detect an input image

Parameters

im – (numpy.ndarray)The input image data, 3-D array with layout HWC, BGR format

Returns

DetectionResult

property preprocessor#

Get PaddleDetPreprocessor object of the loaded model

:return PaddleDetPreprocessor

fastdeploy.vision.detection.FasterRCNN#

class fastdeploy.vision.detection.FasterRCNN(model_file, params_file, config_file, runtime_option=None, model_format=<ModelFormat.PADDLE: 1>)[source]#

Load a FasterRCNN model exported by PaddleDetection.

Parameters

  • model_file – (str)Path of model file, e.g fasterrcnn/model.pdmodel

  • params_file – (str)Path of parameters file, e.g fasterrcnn/model.pdiparams, if the model_fomat is ModelFormat.ONNX, this param will be ignored, can be set as empty string

  • config_file – (str)Path of configuration file for deployment, e.g ppyoloe/infer_cfg.yml

  • runtime_option – (fastdeploy.RuntimeOption)RuntimeOption for inference this model, if it’s None, will use the default backend on CPU

  • model_format – (fastdeploy.ModelForamt)Model format of the loaded model

batch_predict(images)#

Detect a batch of input image list

Parameters

im – (list of numpy.ndarray) The input image list, each element is a 3-D array with layout HWC, BGR format

:return list of DetectionResult

clone()[source]#

Clone FasterRCNN object

Returns

a new FasterRCNN object

get_profile_time()#

Get profile time of Runtime after the profile process is done.

property postprocessor#

Get PaddleDetPostprocessor object of the loaded model

:return PaddleDetPostprocessor

predict(im)#

Detect an input image

Parameters

im – (numpy.ndarray)The input image data, 3-D array with layout HWC, BGR format

Returns

DetectionResult

property preprocessor#

Get PaddleDetPreprocessor object of the loaded model

:return PaddleDetPreprocessor

fastdeploy.vision.detection.MaskRCNN#

class fastdeploy.vision.detection.MaskRCNN(model_file, params_file, config_file, runtime_option=None, model_format=<ModelFormat.PADDLE: 1>)[source]#

Load a MaskRCNN model exported by PaddleDetection.

Parameters

  • model_file – (str)Path of model file, e.g fasterrcnn/model.pdmodel

  • params_file – (str)Path of parameters file, e.g fasterrcnn/model.pdiparams, if the model_fomat is ModelFormat.ONNX, this param will be ignored, can be set as empty string

  • config_file – (str)Path of configuration file for deployment, e.g ppyoloe/infer_cfg.yml

  • runtime_option – (fastdeploy.RuntimeOption)RuntimeOption for inference this model, if it’s None, will use the default backend on CPU

  • model_format – (fastdeploy.ModelForamt)Model format of the loaded model

batch_predict(images)[source]#

Detect a batch of input image list, batch_predict is not supported for maskrcnn now.

Parameters

im – (list of numpy.ndarray) The input image list, each element is a 3-D array with layout HWC, BGR format

:return list of DetectionResult

clone()[source]#

Clone MaskRCNN object

Returns

a new MaskRCNN object

get_profile_time()#

Get profile time of Runtime after the profile process is done.

property postprocessor#

Get PaddleDetPostprocessor object of the loaded model

:return PaddleDetPostprocessor

predict(im)#

Detect an input image

Parameters

im – (numpy.ndarray)The input image data, 3-D array with layout HWC, BGR format

Returns

DetectionResult

property preprocessor#

Get PaddleDetPreprocessor object of the loaded model

:return PaddleDetPreprocessor

fastdeploy.vision.detection.NanoDetPlus#

class fastdeploy.vision.detection.NanoDetPlus(model_file, params_file='', runtime_option=None, model_format=<ModelFormat.ONNX: 2>)[source]#

Load a NanoDetPlus model exported by NanoDet.

Parameters

  • model_file – (str)Path of model file, e.g ./nanodet.onnx

  • params_file – (str)Path of parameters file, e.g yolox/model.pdiparams, if the model_fomat is ModelFormat.ONNX, this param will be ignored, can be set as empty string

  • runtime_option – (fastdeploy.RuntimeOption)RuntimeOption for inference this model, if it’s None, will use the default backend on CPU

  • model_format – (fastdeploy.ModelForamt)Model format of the loaded model

get_profile_time()#

Get profile time of Runtime after the profile process is done.

predict(input_image, conf_threshold=0.25, nms_iou_threshold=0.5)[source]#

Detect an input image

Parameters

  • input_image – (numpy.ndarray)The input image data, 3-D array with layout HWC, BGR format

  • conf_threshold – confidence threashold for postprocessing, default is 0.25

  • nms_iou_threshold – iou threashold for NMS, default is 0.5

Returns

DetectionResult

property reg_max#

reg_max for GFL regression, default 7

property size#

Argument for image preprocessing step, the preprocess image size, tuple of (width, height), default (320, 320)

fastdeploy.vision.detection.ScaledYOLOv4#

class fastdeploy.vision.detection.ScaledYOLOv4(model_file, params_file='', runtime_option=None, model_format=<ModelFormat.ONNX: 2>)[source]#

Load a ScaledYOLOv4 model exported by ScaledYOLOv4.

Parameters

  • model_file – (str)Path of model file, e.g ./scaled_yolov4.onnx

  • params_file – (str)Path of parameters file, e.g yolox/model.pdiparams, if the model_fomat is ModelFormat.ONNX, this param will be ignored, can be set as empty string

  • runtime_option – (fastdeploy.RuntimeOption)RuntimeOption for inference this model, if it’s None, will use the default backend on CPU

  • model_format – (fastdeploy.ModelForamt)Model format of the loaded model

get_profile_time()#

Get profile time of Runtime after the profile process is done.

predict(input_image, conf_threshold=0.25, nms_iou_threshold=0.5)[source]#

Detect an input image

Parameters

  • input_image – (numpy.ndarray)The input image data, 3-D array with layout HWC, BGR format

  • conf_threshold – confidence threashold for postprocessing, default is 0.25

  • nms_iou_threshold – iou threashold for NMS, default is 0.5

Returns

DetectionResult

property size#

Argument for image preprocessing step, the preprocess image size, tuple of (width, height), default size = [640, 640]

fastdeploy.vision.detection.YOLOR#

class fastdeploy.vision.detection.YOLOR(model_file, params_file='', runtime_option=None, model_format=<ModelFormat.ONNX: 2>)[source]#

Load a YOLOR model exported by YOLOR

Parameters

  • model_file – (str)Path of model file, e.g ./yolor.onnx

  • params_file – (str)Path of parameters file, e.g yolox/model.pdiparams, if the model_fomat is ModelFormat.ONNX, this param will be ignored, can be set as empty string

  • runtime_option – (fastdeploy.RuntimeOption)RuntimeOption for inference this model, if it’s None, will use the default backend on CPU

  • model_format – (fastdeploy.ModelForamt)Model format of the loaded model

get_profile_time()#

Get profile time of Runtime after the profile process is done.

predict(input_image, conf_threshold=0.25, nms_iou_threshold=0.5)[source]#

Detect an input image

Parameters

  • input_image – (numpy.ndarray)The input image data, 3-D array with layout HWC, BGR format

  • conf_threshold – confidence threashold for postprocessing, default is 0.25

  • nms_iou_threshold – iou threashold for NMS, default is 0.5

Returns

DetectionResult

property size#

Argument for image preprocessing step, the preprocess image size, tuple of (width, height), default size = [640, 640]

fastdeploy.vision.detection.YOLOv5Preprocessor#

class fastdeploy.vision.detection.YOLOv5Preprocessor[source]#

Create a preprocessor for YOLOv5

property is_mini_pad#

is_mini_pad for preprocessing, pad to the minimum rectange which height and width is times of stride, default false

property is_scale_up#

is_scale_up for preprocessing, the input image only can be zoom out, the maximum resize scale cannot exceed 1.0, default true

property padding_value#

padding value for preprocessing, default [114.0, 114.0, 114.0]

run(input_ims)[source]#

Preprocess input images for YOLOv5

Param

input_ims: (list of numpy.ndarray)The input image

Returns

list of FDTensor

property size#

Argument for image preprocessing step, the preprocess image size, tuple of (width, height), default size = [640, 640]

property stride#

stride for preprocessing, only for mini_pad mode, default 32

fastdeploy.vision.detection.YOLOv5Postprocessor#

class fastdeploy.vision.detection.YOLOv5Postprocessor[source]#

Create a postprocessor for YOLOv5

property conf_threshold#

confidence threshold for postprocessing, default is 0.25

property multi_label#

multi_label for postprocessing, set true for eval, default is True

property nms_threshold#

nms threshold for postprocessing, default is 0.5

run(runtime_results, ims_info)[source]#

Postprocess the runtime results for YOLOv5

Param

runtime_results: (list of FDTensor)The output FDTensor results from runtime

Param

ims_info: (list of dict)Record input_shape and output_shape

Returns

list of DetectionResult(If the runtime_results is predict by batched samples, the length of this list equals to the batch size)

fastdeploy.vision.detection.YOLOv5#

class fastdeploy.vision.detection.YOLOv5(model_file, params_file='', runtime_option=None, model_format=<ModelFormat.ONNX: 2>)[source]#

Load a YOLOv5 model exported by YOLOv5.

Parameters

  • model_file – (str)Path of model file, e.g ./yolov5.onnx

  • params_file – (str)Path of parameters file, e.g yolox/model.pdiparams, if the model_fomat is ModelFormat.ONNX, this param will be ignored, can be set as empty string

  • runtime_option – (fastdeploy.RuntimeOption)RuntimeOption for inference this model, if it’s None, will use the default backend on CPU

  • model_format – (fastdeploy.ModelForamt)Model format of the loaded model

batch_predict(images)[source]#

Classify a batch of input image

Parameters

im – (list of numpy.ndarray) The input image list, each element is a 3-D array with layout HWC, BGR format

:return list of DetectionResult

get_profile_time()#

Get profile time of Runtime after the profile process is done.

property postprocessor#

Get YOLOv5Postprocessor object of the loaded model

:return YOLOv5Postprocessor

predict(input_image, conf_threshold=0.25, nms_iou_threshold=0.5)[source]#

Detect an input image

Parameters

  • input_image – (numpy.ndarray)The input image data, 3-D array with layout HWC, BGR format

  • conf_threshold – confidence threshold for postprocessing, default is 0.25

  • nms_iou_threshold – iou threshold for NMS, default is 0.5

Returns

DetectionResult

property preprocessor#

Get YOLOv5Preprocessor object of the loaded model

:return YOLOv5Preprocessor

fastdeploy.vision.detection.YOLOv5Lite#

class fastdeploy.vision.detection.YOLOv5Lite(model_file, params_file='', runtime_option=None, model_format=<ModelFormat.ONNX: 2>)[source]#

Load a YOLOv5Lite model exported by YOLOv5Lite.

Parameters

  • model_file – (str)Path of model file, e.g ./yolov5lite.onnx

  • params_file – (str)Path of parameters file, e.g yolox/model.pdiparams, if the model_fomat is ModelFormat.ONNX, this param will be ignored, can be set as empty string

  • runtime_option – (fastdeploy.RuntimeOption)RuntimeOption for inference this model, if it’s None, will use the default backend on CPU

  • model_format – (fastdeploy.ModelForamt)Model format of the loaded model

property downsample_strides#

downsample strides for YOLOv5Lite to generate anchors, will take (8,16,32) as default values, might have stride=64.

get_profile_time()#

Get profile time of Runtime after the profile process is done.

property is_decode_exported#

whether the model_file was exported with decode module. The official YOLOv5Lite/export.py script will export ONNX file without decode module. Please set it ‘true’ manually if the model file was exported with decode module. False : ONNX files without decode module. True : ONNX file with decode module. default False

predict(input_image, conf_threshold=0.25, nms_iou_threshold=0.5)[source]#

Detect an input image

Parameters

  • input_image – (numpy.ndarray)The input image data, 3-D array with layout HWC, BGR format

  • conf_threshold – confidence threashold for postprocessing, default is 0.25

  • nms_iou_threshold – iou threashold for NMS, default is 0.5

Returns

DetectionResult

property size#

Argument for image preprocessing step, the preprocess image size, tuple of (width, height), default size = [640, 640]

fastdeploy.vision.detection.YOLOv6#

class fastdeploy.vision.detection.YOLOv6(model_file, params_file='', runtime_option=None, model_format=<ModelFormat.ONNX: 2>)[source]#

Load a YOLOv6 model exported by YOLOv6.

Parameters

  • model_file – (str)Path of model file, e.g ./yolov6.onnx

  • params_file – (str)Path of parameters file, e.g yolox/model.pdiparams, if the model_fomat is ModelFormat.ONNX, this param will be ignored, can be set as empty string

  • runtime_option – (fastdeploy.RuntimeOption)RuntimeOption for inference this model, if it’s None, will use the default backend on CPU

  • model_format – (fastdeploy.ModelForamt)Model format of the loaded model

get_profile_time()#

Get profile time of Runtime after the profile process is done.

predict(input_image, conf_threshold=0.25, nms_iou_threshold=0.5)[source]#

Detect an input image

Parameters

  • input_image – (numpy.ndarray)The input image data, 3-D array with layout HWC, BGR format

  • conf_threshold – confidence threashold for postprocessing, default is 0.25

  • nms_iou_threshold – iou threashold for NMS, default is 0.5

Returns

DetectionResult

property size#

Argument for image preprocessing step, the preprocess image size, tuple of (width, height), default size = [640, 640]

fastdeploy.vision.detection.YOLOv7Preprocessor#

class fastdeploy.vision.detection.YOLOv7Preprocessor[source]#

Create a preprocessor for YOLOv7

property is_scale_up#

is_scale_up for preprocessing, the input image only can be zoom out, the maximum resize scale cannot exceed 1.0, default true

property padding_value#

padding value for preprocessing, default [114.0, 114.0, 114.0]

run(input_ims)[source]#

Preprocess input images for YOLOv7

Param

input_ims: (list of numpy.ndarray)The input image

Returns

list of FDTensor

property size#

Argument for image preprocessing step, the preprocess image size, tuple of (width, height), default size = [640, 640]

fastdeploy.vision.detection.YOLOv7Postprocessor#

class fastdeploy.vision.detection.YOLOv7Postprocessor[source]#

Create a postprocessor for YOLOv7

property conf_threshold#

confidence threshold for postprocessing, default is 0.25

property nms_threshold#

nms threshold for postprocessing, default is 0.5

run(runtime_results, ims_info)[source]#

Postprocess the runtime results for YOLOv7

Param

runtime_results: (list of FDTensor)The output FDTensor results from runtime

Param

ims_info: (list of dict)Record input_shape and output_shape

Returns

list of DetectionResult(If the runtime_results is predict by batched samples, the length of this list equals to the batch size)

fastdeploy.vision.detection.YOLOv7#

class fastdeploy.vision.detection.YOLOv7(model_file, params_file='', runtime_option=None, model_format=<ModelFormat.ONNX: 2>)[source]#

Load a YOLOv7 model exported by YOLOv7.

Parameters

  • model_file – (str)Path of model file, e.g ./yolov7.onnx

  • params_file – (str)Path of parameters file, e.g yolox/model.pdiparams, if the model_fomat is ModelFormat.ONNX, this param will be ignored, can be set as empty string

  • runtime_option – (fastdeploy.RuntimeOption)RuntimeOption for inference this model, if it’s None, will use the default backend on CPU

  • model_format – (fastdeploy.ModelForamt)Model format of the loaded model

batch_predict(images)[source]#

Classify a batch of input image

Parameters

im – (list of numpy.ndarray) The input image list, each element is a 3-D array with layout HWC, BGR format

:return list of DetectionResult

get_profile_time()#

Get profile time of Runtime after the profile process is done.

property postprocessor#

Get YOLOv7Postprocessor object of the loaded model

:return YOLOv7Postprocessor

predict(input_image, conf_threshold=0.25, nms_iou_threshold=0.5)[source]#

Detect an input image

Parameters

  • input_image – (numpy.ndarray)The input image data, 3-D array with layout HWC, BGR format

  • conf_threshold – confidence threshold for postprocessing, default is 0.25

  • nms_iou_threshold – iou threshold for NMS, default is 0.5

Returns

DetectionResult

property preprocessor#

Get YOLOv7Preprocessor object of the loaded model

:return YOLOv7Preprocessor

fastdeploy.vision.detection.YOLOR#

class fastdeploy.vision.detection.YOLOR(model_file, params_file='', runtime_option=None, model_format=<ModelFormat.ONNX: 2>)[source]#

Load a YOLOR model exported by YOLOR

Parameters

  • model_file – (str)Path of model file, e.g ./yolor.onnx

  • params_file – (str)Path of parameters file, e.g yolox/model.pdiparams, if the model_fomat is ModelFormat.ONNX, this param will be ignored, can be set as empty string

  • runtime_option – (fastdeploy.RuntimeOption)RuntimeOption for inference this model, if it’s None, will use the default backend on CPU

  • model_format – (fastdeploy.ModelForamt)Model format of the loaded model

get_profile_time()#

Get profile time of Runtime after the profile process is done.

predict(input_image, conf_threshold=0.25, nms_iou_threshold=0.5)[source]#

Detect an input image

Parameters

  • input_image – (numpy.ndarray)The input image data, 3-D array with layout HWC, BGR format

  • conf_threshold – confidence threashold for postprocessing, default is 0.25

  • nms_iou_threshold – iou threashold for NMS, default is 0.5

Returns

DetectionResult

property size#

Argument for image preprocessing step, the preprocess image size, tuple of (width, height), default size = [640, 640]

fastdeploy.vision.detection.YOLOv7End2EndORT#

class fastdeploy.vision.detection.YOLOv7End2EndORT(model_file, params_file='', runtime_option=None, model_format=<ModelFormat.ONNX: 2>)[source]#

Load a YOLOv7End2EndORT model exported by YOLOv7.

Parameters

  • model_file – (str)Path of model file, e.g ./yolov7end2end_ort.onnx

  • params_file – (str)Path of parameters file, e.g yolox/model.pdiparams, if the model_fomat is ModelFormat.ONNX, this param will be ignored, can be set as empty string

  • runtime_option – (fastdeploy.RuntimeOption)RuntimeOption for inference this model, if it’s None, will use the default backend on CPU

  • model_format – (fastdeploy.ModelForamt)Model format of the loaded model

get_profile_time()#

Get profile time of Runtime after the profile process is done.

predict(input_image, conf_threshold=0.25)[source]#

Detect an input image

Parameters

  • input_image – (numpy.ndarray)The input image data, 3-D array with layout HWC, BGR format

  • conf_threshold – confidence threashold for postprocessing, default is 0.25

Returns

DetectionResult

property size#

Argument for image preprocessing step, the preprocess image size, tuple of (width, height), default size = [640, 640]

fastdeploy.vision.detection.YOLOv7End2EndTRT#

class fastdeploy.vision.detection.YOLOv7End2EndTRT(model_file, params_file='', runtime_option=None, model_format=<ModelFormat.ONNX: 2>)[source]#

Load a YOLOv7End2EndTRT model exported by YOLOv7.

Parameters

  • model_file – (str)Path of model file, e.g ./yolov7end2end_trt.onnx

  • params_file – (str)Path of parameters file, e.g yolox/model.pdiparams, if the model_fomat is ModelFormat.ONNX, this param will be ignored, can be set as empty string

  • runtime_option – (fastdeploy.RuntimeOption)RuntimeOption for inference this model, if it’s None, will use the default backend on CPU

  • model_format – (fastdeploy.ModelForamt)Model format of the loaded model

get_profile_time()#

Get profile time of Runtime after the profile process is done.

predict(input_image, conf_threshold=0.25)[source]#

Detect an input image

Parameters

  • input_image – (numpy.ndarray)The input image data, 3-D array with layout HWC, BGR format

  • conf_threshold – confidence threashold for postprocessing, default is 0.25

Returns

DetectionResult

property size#

Argument for image preprocessing step, the preprocess image size, tuple of (width, height), default size = [640, 640]

fastdeploy.vision.detection.YOLOX#

class fastdeploy.vision.detection.YOLOX(model_file, params_file='', runtime_option=None, model_format=<ModelFormat.ONNX: 2>)[source]#

Load a YOLOX model exported by YOLOX.

Parameters

  • model_file – (str)Path of model file, e.g ./yolox.onnx

  • params_file – (str)Path of parameters file, e.g yolox/model.pdiparams, if the model_fomat is ModelFormat.ONNX, this param will be ignored, can be set as empty string

  • runtime_option – (fastdeploy.RuntimeOption)RuntimeOption for inference this model, if it’s None, will use the default backend on CPU

  • model_format – (fastdeploy.ModelForamt)Model format of the loaded model

property downsample_strides#

downsample strides for YOLOX to generate anchors, will take (8,16,32) as default values, might have stride=64.

get_profile_time()#

Get profile time of Runtime after the profile process is done.

property is_decode_exported#

whether the model_file was exported with decode module. The official YOLOX/tools/export_onnx.py script will export ONNX file without decode module. Please set it ‘true’ manually if the model file was exported with decode module. Defalut False.

predict(input_image, conf_threshold=0.25, nms_iou_threshold=0.5)[source]#

Detect an input image

Parameters

  • input_image – (numpy.ndarray)The input image data, 3-D array with layout HWC, BGR format

  • conf_threshold – confidence threashold for postprocessing, default is 0.25

  • nms_iou_threshold – iou threashold for NMS, default is 0.5

Returns

DetectionResult

property size#

Argument for image preprocessing step, the preprocess image size, tuple of (width, height), default size = [640, 640]