A Multi-Task Learning and Data Augmentation-Based Pose Estimation Algorithm
Hui Li, Jiangyuan Qi · 2023
Pose estimation is a fundamental task in computer vision that aims to localize human body joints in images or videos. However, the performance of pose estimation algorithms can be limited by various factors such as occlusions, complex body poses, and variations in clothing and lighting. To address these challenges, we propose a novel pose estimation algorithm based on multi-task learning and data augmentation. Our method utilizes a multi-task learning framework to incorporate multiple datasets as training data, leading to improved learning performance. We also employ various data augmentation techniques to improve the generalization ability of the network and mitigate the effects of data imbalance. Experimental results on the COCO dataset and MPII Human Pose dataset demonstrate that our method has higher accuracy and stronger robustness compared to the baseline algorithm on pose estimation tasks.