Head Pose Estimation

Overviews

Head Pose Estimation

The packages contain 2 different algorithms:

1. 6DRepNet (sixd_repnet)

   Original git(6DRepNet): https://github.com/thohemp/6DRepNet.git

2. DirectMHP

   Original git(DirectMHP): https://github.com/hnuzhy/DirectMHP.git

Example DirectMHP

In this document, we will show you how to use HACHIX version of these algorithms.

APIs

Model

Bases: BaseModel

build_model(custom_weight=None)

Build model. Type of model is defined in opt.json. Can be "sixd_repnet" or "directmhp"

Parameters:

Name	Type	Description	Default
custom_weight	str	Custom weight. Defaults to None.	None

Returns:

Name	Type	Description
model	Model	model sixd_repnet or model directmhp depends on opt.json

fit()

Train model

get_transform()

Transform input by ImageNet normalization.

Returns:

Name	Type	Description
Compose		A Transform Compose

predict(img_path)

Predict using Head Pose Estimation

Parameters:

Name	Type	Description	Default
img_path	str	The path to the input image file.	required

With model type is sixd_repnet Returns: img (np.array): The output image. pitchs (List): Predicted pitch values. yaws (List): Predicted yaw values. rolls (List): Predicted roll values.

With model type is directmhp Returns: img0 (np.array): The processed image. bboxes (numpy.ndarray): Predicted bounding boxes. scores (numpy.ndarray): Predicted scores. pitchs_yaws_rolls (numpy.ndarray): Predicted pitch, yaw, and roll values.

process_predictions(net_output, raw_image_path, raw_image_mat, image_transform, save_image_path)

post process the output of the net

Parameters:

Name	Type	Description	Default
net_output	_type_	output of detection net	required
raw_image_path	str	raw image path	required
raw_image_mat	Array	raw image numpy mat type	required
image_transform	func	image_transform function	required
save_image_path	str	save image path	required

Returns:

Name	Type	Description
save_image_path	str	inform that the annotation image has been written successfully