Adding metadata to ONNX model

The plugin allows you to load the meta parameters of the onnx model automatically. Predefined parameter types are designed for the simplicity of the user’s work. The written metadata is especially important for users who are not familiar with the technical details.

List of parameters parsed by plugin

Parameter | Type | Example | Description
model_type | str | 'Segmentor' | Types of models.
class_names		dict		{0: 'name1', 1: 'name2'}		A dictionary that maps a class id to its name.
class_names (mutli outputs)		list		[{0: 'name1'}, {0: 'name2'}]		A dictionary that maps a class id to its name.
resolution		float		100		Real-world resolution of images (centimeters per pixel).
tiles_size		int		512		What size (in pixels) is the tile to crop.
tiles_overlap		int		40		How many percent of the image size overlap.
standardization_mean		list		[0.0, 0.0, 0.0]		Mean - if you want to standarize input after normalisation
standardization_std		list		[1.0, 1.0, 1.0]		Std - if you want to standarize input after normalisation
seg_thresh		float		0.5		Segmentor: class confidence threshold.
seg_small_segment		int		7		Segmentor: remove small occurrences of the class.
reg_output_scaling		float		1.0		Regressor: scaling factor for the model output.
det_conf		float		0.6		Detector: object confidence threshold.
det_iou_thresh		float		0.4		Detector: IOU threshold for NMS.
det_type		str		YOLO_v5_or_v7_default		Detector: type of the detector model format

Available model types: - Segmentor - Detector - Regressor - Recognition - Superresolution

Availeble detector types: - YOLO_v5_or_v7_default - YOLO_v6 - YOLO_v9 - YOLO_Ultralytics - YOLO_Ultralytics_segmentation

Example

The example below shows how to add string, float, and dictionary metadata into a model. Note that metadata is created while model.metadata_props.add() is called. Moreover, the metadata value has to be a byte type.

import json
import onnx

model = onnx.load('deeplabv3_landcover_4c.onnx')

class_names = {
    0: '_background',
    1: 'building',
    2: 'woodland',
    3: 'water',
    4: 'road',
}

m1 = model.metadata_props.add()
m1.key = 'model_type'
m1.value = json.dumps('Segmentor')

m2 = model.metadata_props.add()
m2.key = 'class_names'
m2.value = json.dumps(class_names)

m3 = model.metadata_props.add()
m3.key = 'resolution'
m3.value = json.dumps(50)

# optional, if you want to standarize input after normalisation
m4 = model.metadata_props.add()
m4.key = 'standardization_mean'
m4.value = json.dumps([0.0, 0.0, 0.0])

m5 = model.metadata_props.add()
m5.key = 'standardization_std'
m5.value = json.dumps([1.0, 1.0, 1.0])

onnx.save(model, 'deeplabv3_landcover_4c.onnx')

You can also use a script setting the most important parameters in the plugin repository (file tools/add_model_metadata.py).