modify
add_poseflow_figures
add_poseflow_figures (input_detections, json_path)
output_alphapose_json
output_alphapose_json (poses_series, figure_type='figures')
interpolate_missing_coords
interpolate_missing_coords (input_frames, threshold=0.5, flip_figures=False, check_bbox=False, all_visible=True, overlap_threshold=0.7, video_file=None, figure_type='figures')
trim_empty_frames_start_end
trim_empty_frames_start_end (input_frames, figure_type='figures')
interpolate_missing_poses
interpolate_missing_poses (input_frames, threshold=0.2, video_file=None, figure_type='figures', trim_ends=True)
is_usable_pose
is_usable_pose (frame_info, threshold=0.2, figure_type='figures')
add_flipped_zeroified_figures
add_flipped_zeroified_figures (input_frames, add_flipped=True, add_zerofied=True, figure_type='figures')
correct_pose
correct_pose (input_coords)
right_ankle_from_knee
right_ankle_from_knee (coords, missing_coords)
left_ankle_from_knee
left_ankle_from_knee (coords, missing_coords)
right_hip_btwn_shoulder_knee_ankle
right_hip_btwn_shoulder_knee_ankle (coords, missing_coords)
left_hip_btwn_shoulder_knee_ankle
left_hip_btwn_shoulder_knee_ankle (coords, missing_coords)
right_elbow_btwn_shoulder_wrist
right_elbow_btwn_shoulder_wrist (coords, missing_coords)
left_elbow_btwn_shoulder_wrist
left_elbow_btwn_shoulder_wrist (coords, missing_coords)
right_ear_btwn_eye_shoulder
right_ear_btwn_eye_shoulder (coords, missing_coords)
left_ear_btwn_eye_shoulder
left_ear_btwn_eye_shoulder (coords, missing_coords)
right_eye_btwn_nose_shoulder
right_eye_btwn_nose_shoulder (coords, missing_coords)
left_eye_btwn_nose_shoulder
left_eye_btwn_nose_shoulder (coords, missing_coords)
nose_btwn_eyes_ears_shoulders
nose_btwn_eyes_ears_shoulders (coords, missing_coords)
count_figures_and_time
count_figures_and_time (input_frames, figure_type='figures')
average_coords
average_coords (coord1, coord2)
shift_figure
shift_figure (coords_and_confidence, dx, dy)
get_union
get_union (a, b)
get_intersect
get_intersect (a, b)
in_bbox_check
in_bbox_check (coord, bbox, margin=0.5)
get_bbox_area
get_bbox_area (bbox)
get_bbox
get_bbox (pose_coords, move_to_origin=False, margin=0, width=None, height=None)
zeroify_detections_y_first
zeroify_detections_y_first (input_detections, width=None, height=None)
Version of zeroify_detections from above that assumes Y, X order of coordinates, again with Y=0 at top left
zeroify_detections
zeroify_detections (input_detections, width=None, height=None)
Modifies a figure’s coordinates so that the corner of its bounding box is at 0,0. This is mostly for visualization with PIL images, (note that PIL puts y=0 at the top). The modifications are done for all figures in a single frame.
flip_detections_y_first
flip_detections_y_first (input_detections, flip_y=False, flip_x=False, rectify_x=False, mirror_coco_17_left_right=False)
Mirror the coordinates of a pose in place around the midpoint of either the Y or X axis. The former is sometimes necessary when working with image coordinates (0,0 at top left) vs. figure coordiantes (0,0 at bottom left). The latter is useful when the pose needs to be flipped as though it is being viewed in a mirror (which is often the case with dance videos).
This function flips all detected poses in a single frame. If the rectify_x flag is set, this function can also count how many coords are on the left or right side of the pose, and mirror the coordinates horizontally so that the most coords are always on stage right (viewer’s left) – in which case the parameter value of flip_x is ignored.
flip_detections
flip_detections (input_detections, flip_y=False, flip_x=False, rectify_x=False, y_first=False, mirror_coco_17_left_right=False)
Mirror the coordinates of a pose in place around the midpoint of either the Y or X axis. The former is sometimes necessary when working with image coordinates (0,0 at top left) vs. figure coordiantes (0,0 at bottom left). The latter is useful when the pose needs to be flipped as though it is being viewed in a mirror (which is often the case with dance videos).
This function flips all detected poses in a single frame. If the rectify_x flag is set, this function can also count how many coords are on the left or right side of the pose, and mirror the coordinates horizontally so that the most coords are always on stage right (viewer’s left) – in which case the parameter value of flip_x is ignored.
get_figure_coords_y_first
get_figure_coords_y_first (coords_and_confidence, margin=0)
get_figure_coords
get_figure_coords (coords_and_confidence, margin=0)