Toward Reliable Infant Pose Estimation: A Training-Dynamics Approach to Noisy Annotation Detection

2026-10-05

AI Summary

This study addresses the distortion of clinical motor biomarkers in premature infant pose estimation caused by noisy manually annotated keypoints. To mitigate this issue, it introduces the "small-loss" hypothesis into keypoint noise detection for the first time, proposing an unsupervised filtering framework based on cross-entropy training dynamics. Specifically, the framework integrates a hybrid convolutional attention network with clustering algorithms to leverage training dynamics for predicting label categories and separating clean from noisy data. Evaluated on the NeoPose dataset, the proposed method achieves an F1 score of 91.9% and yields a 7.4 AP improvement in downstream accuracy on the COCO benchmark. These results demonstrate that this work provides a robust noisy-label learning solution for neonatal pose analysis.

Abstract

Spontaneous movement analysis in preterm infants relies increasingly on markerless pose estimation (PE) to derive clinically relevant motion biomarkers directly from video recordings. Training accurate infant PE models requires large sets of manually annotated keypoints, and human annotation is inherently prone to error. Noisy keypoints (i.e., keypoints mislocalized with respect to their true anatomical position) are especially problematic in this clinical setting, since they can propagate as artificial artifacts into the reconstructed joint trajectories. Building on the small-loss hypothesis and training-dynamics-based sample selection established in the noisy-label learning literature, we propose a novel framework for detecting noisy keypoint annotations. A hybrid convolutional-attention model is trained to predict the anatomical category of each keypoint from its spatial coordinates and local visual features; the resulting cross-entropy training dynamics are then used to derive per-keypoint descriptors, which are partitioned into clean and noisy subsets via unsupervised clustering. We validate the approach on NeoPose, a newly collected dataset of 65 hospitalized preterm infants, under two realistic noise scenarios (random positional perturbation and left-right swapping) across multiple noise levels. Results show that the proposed approach achieves an F1-score of up to 91.9% in noisy-keypoint detection. The framework further generalizes to the heterogeneous COCO benchmark, where filtering CE-detected noisy keypoints from the training set also yields measurable improvements (up to 7.4 AP points) in downstream pose estimation accuracy at moderate-to-high noise levels.

Problem

• Infant Pose Estimation
• Noisy Annotation Detection
• Keypoint Localization
• Training Dynamics

Innovation

• Noisy Annotation Detection
• Training Dynamics
• Infant Pose Estimation
• Hybrid Convolutional-Attention Model
• Unsupervised Clustering