VitPose

Overview

Simple Vision Transformer Baselines for Human Pose Estimation

APIs

Model

Bases: BaseModel

build_model(det_model_name, pose_model_name)

Load model from pre-training

Parameters:

Name	Type	Description	Default
det_model_name	str	detection model name	required
pose_model_name	str	pose model name	required

Returns:

Name	Type	Description
_type_		description

process_predictions(video_path, box_score_threshold, max_num_frames, kpt_score_threshold, vis_dot_radius, vis_line_thickness)

process video

Parameters:

Name	Type	Description	Default
video_path	str	video path	required
box_score_threshold	float	score threshold for box	required
max_num_frames	int	maximum number of frames	required
kpt_score_threshold	float	KPT score threshold	required
vis_dot_radius	int	radius of point	required
vis_line_thickness	int	line thickness	required

Returns:

Type	Description
tuple[str, list[list[dict[str, ndarray]]]]	tuple[str, list[list[dict[str, np.ndarray]]]]: description