One, Five, and Ten-Shot-Based Meta-Learning for Computationally Efficient Head Pose Estimation

• Published in: International journal of embedded and real-time communication systems Vol. 14; no. 1; pp. 1 - 24
• Main Authors: Joshi, Manoj, Pant, Dibakar Raj, Heikkonen, Jukka, Kanth, Rajeev
• Format: Journal Article
• Language: English
• Published: Hershey IGI Global 01.01.2023
• Subjects: Artificial neural networks; Communication; Computational efficiency; Datasets; Deep learning; Efficiency; Learning; Methods; Neural networks; Pose estimation; Sensors; Surveillance
• ISSN: 1947-3176; 1947-3184; 1947-3184
• DOI: 10.4018/IJERTCS.316877

Many real-world applications rely on head pose estimation. The performance of head pose estimation has significantly improved with techniques like convolutional neural networks (CNN). However, CNN requires a large amount of data for training. This article presents a new framework for head pose estimation using computationally efficient first-order model-agnostic meta-learning (FO-MAML)-based method and compares the performance with existing MAML-based approaches. Experiments using one-shot, five-shot, and ten-shot settings are done using MAML and FO-MAML. A mean average error (MAEavg) of 7.72, 6.30, and 5.32 has been achieved in predicting head pose using MAML for one-, five-, and ten-shot settings, respectively. Similarly, MAEavg of 8.33, 6.84, and 6.23 has been achieved in predicting head pose using FO-MAML for one-, five-, and ten-shot settings, respectively. The computational complexity of an outer-loop update in MAML is found to be O(n2) whereas for FO-MAML it is O(n).