Human Parsing
BiWAKO.HumanParsing
Basic ResNet50 model for parsing attributes of pedestrians
Model
This model is trained on the pedestrian dataset with following multi-label classification task: - "is_male", - "has_bag", - "has_hat", - "has_longsleeves", - "has_longpants", - "has_longhair", - "has_coat_jacket",
Attributes:
| Name | Type | Description |
|---|---|---|
model |
onnxruntime.InferenceSession |
ONNXRuntime instance. |
conf_thresh |
float |
confidence threshold for prediction. |
input_name |
str |
name of input node. |
output_name |
str |
name of output node. |
input_shape |
tuple |
shape of input node. |
labels |
np.ndarray |
mapping of label index to label name. |
__init__(self, model='human_attribute', conf_thresh=0.4)
special
Initialize HumanParsing
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
model |
str |
model name or path to the onnx file. Defaults to "human_attribute". |
'human_attribute' |
conf_thresh |
float |
confidence threshold. Defaults to 0.4. |
0.4 |
predict(self, image)
Return the prediction on the image.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
image |
Image |
image to be processed in str or cv2 format. |
required |
Returns:
| Type | Description |
|---|---|
np.ndarray |
processed image |
render(self, prediction, image)
Return the original image with the prediction at the top left corner.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
prediction |
np.ndarray |
prediction of the model. |
required |
image |
Image |
image to be rendered in str or cv2 format. |
required |
Returns:
| Type | Description |
|---|---|
np.ndarray |
rendered image |