Configuration

Parameter table

Parameter name	Parameter type	Parameter function	Notes
Function	StringArray	Algorithm name	-
Thread	Integer	Number of threads	<=4
Sources	String	Resource Directory	Can fill in a single picture, folder. If it is a video stream, leave it blank
pipelineMode	String	Algorithm mode	-
ModelPath	String	Model Path	-
mean	FloatArray	mean	-
normal	FloatArray	Variance	-
min_sizes	FloatArrayArray	Minimum size	You can leave it blank, if you don’t fill it, you need to configure base_sizes, ratios parameters
scales	FloatArrayArray	Scale factor	-
base_sizes	FloatArrayArray	base size	-
ratios	FloatArray	anchor ratio	-
clip	Boolean	Whether to clip or not	-
input_w	Integer	Model input width	-
input_h	Integer	Model input high	-
input_type	String	Model input type	-
strides	FloatArray	Span	-
score_threshold	Float	Score threshold	-
iou_threshold	Float	IOU threshold	-
variance	FloatArray	Variance	Used when generating the detection frame
preprocess	String	Pre-processing mode	-
postprocess	String	Post-processing model	-
TensorOuputString	StringArray	Algorithm output string	"Meaning": "String name"
input_stream	StringArray	Take the result of an algorithm	"Algorithm name": "The meaning of the stream"
output_stream	StringArray	Take only part of the output	"Meaning": "Part taken"

Sample configuration

{
    "Function": ["FaceDetect", "FaceLandmark"],
    "Thread": 2,
    "Sources": "./images/",
    "FaceDetect": {
        "pipelineMode": "Standard",
        "ModelPath": "",
        "mean": [127.0, 127.0, 127.0],
        "normal": [0.0078125, 0.0078125, 0.0078125],
        "min_sizes": [
            [10.0, 16.0, 24.0],
            [32.0, 48.0],
            [64.0, 96.0],
            [128.0, 192.0, 256.0]],
        "scales": [
            [8.0, 4.0, 2.0, 1.0],
            [8.0, 4.0, 2.0, 1.0],
            [32.0, 16.0, 8.0, 4.0]],
        "base_sizes": [[16], [16], [16]],
        "ratios": [2.5],
        "clip": false,
        "input_w": 160,
        "input_h": 120,
        "input_type": "RGB",
        "strides": [8.0, 16.0, 32.0, 64.0],
        "score_threshold": 0.6,
        "iou_threshold": 0.3,
        "variance": [0.1, 0.2],
        "preprocess": "CreateAnchor",
        "postprocess": "NMS",
        "TensorOuputString": ["scores:score", "boxes:box"]
    },
    "FaceLandmark": {
        "ModelPath": "./model/landmark.tmfile",
        "pipelineMode": "MutiInput",
        "mean": [127.0, 127.0, 127.0],
        "normal": [0.007874, 0.007874, 0.007874],
        "input_w": 160,
        "input_h": 160,
        "input_type": "RGB",
        "input_stream": ["FaceDetect:rect"],
        "output_stream": ["points:212"]
    }
}