postprocess.py File Reference

Postprocessing utilities for pose estimation model outputs. More...

Go to the source code of this file.

Namespaces
	postprocess

Functions
def	postprocess.heatmaps_spe_postprocess (tf.Tensor tensor)
def	postprocess.spe_postprocess (tf.Tensor tensor)
def	postprocess._padded_nms (tf.Tensor tensor, int max_output_size, float iou_threshold, float score_threshold)
def	postprocess.yolo_mpe_postprocess (tf.Tensor tensor, int max_output_size=20, float iou_threshold=0.7, float score_threshold=0.25)
def	postprocess.hand_landmarks_postprocess ([tf.Tensor] tensor)
def	postprocess.head_landmarks_postprocess ([tf.Tensor] tensor)

Detailed Description

Postprocessing utilities for pose estimation model outputs.

For MoveNet (heatmaps_spe): decodes (H/4 x W/4 x K) heatmaps to normalized (x, y, confidence) keypoint coordinates. For YOLOv8 (yolo_mpe): applies Non-Maximum Suppression (NMS) to multi-scale bounding box predictions and decodes keypoints.

Note

Politecnico di Milano, A.Y. 2024-2025. Multidisciplinary Project — Neural Network Deployment on STM32N6 NPU.

Definition in file postprocess.py.