Meta-Learning, Fast Adaptation, and Latent Representation for Head Pose Estimation
Head pose estimation is used in a variety of human computer interface applications, like stare tracking, driving assistance, impaired assistance, and entertainment. Advances in convolutional neural networks have a considerable improvement in the performance of head pose estimation. However, difficul...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
FRUCT
2022-04-01
|
| Series: | Proceedings of the XXth Conference of Open Innovations Association FRUCT |
| Subjects: | |
| Online Access: | https://www.fruct.org/publications/fruct31/files/Jos.pdf |
| _version_ | 1818231636059226112 |
|---|---|
| author | Manoj Joshi Dibakar Raj Pant Rupesh Raj Karn Jukka Heikkonen Rajeev Kanth |
| author_facet | Manoj Joshi Dibakar Raj Pant Rupesh Raj Karn Jukka Heikkonen Rajeev Kanth |
| author_sort | Manoj Joshi |
| collection | DOAJ |
| description | Head pose estimation is used in a variety of human computer interface applications, like stare tracking, driving assistance, impaired assistance, and entertainment. Advances in convolutional neural networks have a considerable improvement in the performance of head pose estimation. However, difficulties in capturing well-labelled head pose data and differences in the facial features of different persons make them difficult to use. This work proposes a meta-learning based technique for head pose estimation problem in BIWI head pose dataset. An approach to learning latent representation of head pose features using variational autoencoder is implemented. Then a fast, adaptable head pose estimator is trained using meta-learning in a few-shot settings. Model agnostic meta-learning (MAML) algorithm has been deployed for training a head pose estimator. Mean Average Error (MAEavg) of 7.33 is achieved in predicting head pose angles in one-shot settings. After meta-training, the optimized model is used to analyze fast adaptation in a test set that has been separated from the BIWI head pose dataset. We begin with the trained networks optimum parameters and optimize the inner loop for quick adaptation. The optimized model can predict accurate head poses using as few as 10 gradient descent steps in the unseen set of tasks sampled from the test set. |
| first_indexed | 2024-12-12T10:53:32Z |
| format | Article |
| id | doaj.art-d9f7bfa58d10458b974fdc09020389ea |
| institution | Directory Open Access Journal |
| issn | 2305-7254 2343-0737 |
| language | English |
| last_indexed | 2024-12-12T10:53:32Z |
| publishDate | 2022-04-01 |
| publisher | FRUCT |
| record_format | Article |
| series | Proceedings of the XXth Conference of Open Innovations Association FRUCT |
| spelling | doaj.art-d9f7bfa58d10458b974fdc09020389ea2022-12-22T00:26:43ZengFRUCTProceedings of the XXth Conference of Open Innovations Association FRUCT2305-72542343-07372022-04-01311717810.23919/FRUCT54823.2022.9770932Meta-Learning, Fast Adaptation, and Latent Representation for Head Pose EstimationManoj Joshi0Dibakar Raj Pant1Rupesh Raj Karn2Jukka Heikkonen3Rajeev Kanth4Institute of Engineering / Nepal, NepalInstitute of Engineering / Nepal, NepalKhalifa University, United Arab EmiratesUniversity of Turku, FinlandSavonia University of Applied Sciences / Kuopio, FinlandHead pose estimation is used in a variety of human computer interface applications, like stare tracking, driving assistance, impaired assistance, and entertainment. Advances in convolutional neural networks have a considerable improvement in the performance of head pose estimation. However, difficulties in capturing well-labelled head pose data and differences in the facial features of different persons make them difficult to use. This work proposes a meta-learning based technique for head pose estimation problem in BIWI head pose dataset. An approach to learning latent representation of head pose features using variational autoencoder is implemented. Then a fast, adaptable head pose estimator is trained using meta-learning in a few-shot settings. Model agnostic meta-learning (MAML) algorithm has been deployed for training a head pose estimator. Mean Average Error (MAEavg) of 7.33 is achieved in predicting head pose angles in one-shot settings. After meta-training, the optimized model is used to analyze fast adaptation in a test set that has been separated from the BIWI head pose dataset. We begin with the trained networks optimum parameters and optimize the inner loop for quick adaptation. The optimized model can predict accurate head poses using as few as 10 gradient descent steps in the unseen set of tasks sampled from the test set.https://www.fruct.org/publications/fruct31/files/Jos.pdfhead pose estimationmeta-learningdeep learningrepresentation learningfew-shot learning |
| spellingShingle | Manoj Joshi Dibakar Raj Pant Rupesh Raj Karn Jukka Heikkonen Rajeev Kanth Meta-Learning, Fast Adaptation, and Latent Representation for Head Pose Estimation Proceedings of the XXth Conference of Open Innovations Association FRUCT head pose estimation meta-learning deep learning representation learning few-shot learning |
| title | Meta-Learning, Fast Adaptation, and Latent Representation for Head Pose Estimation |
| title_full | Meta-Learning, Fast Adaptation, and Latent Representation for Head Pose Estimation |
| title_fullStr | Meta-Learning, Fast Adaptation, and Latent Representation for Head Pose Estimation |
| title_full_unstemmed | Meta-Learning, Fast Adaptation, and Latent Representation for Head Pose Estimation |
| title_short | Meta-Learning, Fast Adaptation, and Latent Representation for Head Pose Estimation |
| title_sort | meta learning fast adaptation and latent representation for head pose estimation |
| topic | head pose estimation meta-learning deep learning representation learning few-shot learning |
| url | https://www.fruct.org/publications/fruct31/files/Jos.pdf |
| work_keys_str_mv | AT manojjoshi metalearningfastadaptationandlatentrepresentationforheadposeestimation AT dibakarrajpant metalearningfastadaptationandlatentrepresentationforheadposeestimation AT rupeshrajkarn metalearningfastadaptationandlatentrepresentationforheadposeestimation AT jukkaheikkonen metalearningfastadaptationandlatentrepresentationforheadposeestimation AT rajeevkanth metalearningfastadaptationandlatentrepresentationforheadposeestimation |