Emotion recognition on edge devices: training and deployment
Emotion recognition, among other natural language processing tasks, has greatly benefited from the use of large transformer models. Deploying these models on resource-constrained devices, however, is a major challenge due to their computational cost. In this paper, we show that the combination of la...
| Main Authors: | , , , , |
|---|---|
| Other Authors: | |
| Format: | Journal Article |
| Language: | English |
| Published: |
2022
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/153915 |
| _version_ | 1811696114576916480 |
|---|---|
| author | Pandelea, Vlad Ragusa, Edoardo Apicella, Tommaso Gastaldo, Paolo Cambria, Erik |
| author2 | School of Computer Science and Engineering |
| author_facet | School of Computer Science and Engineering Pandelea, Vlad Ragusa, Edoardo Apicella, Tommaso Gastaldo, Paolo Cambria, Erik |
| author_sort | Pandelea, Vlad |
| collection | NTU |
| description | Emotion recognition, among other natural language processing tasks, has greatly benefited from the use of large transformer models. Deploying these models on resource-constrained devices, however, is a major challenge due to their computational cost. In this paper, we show that the combination of large transformers, as high-quality feature extractors, and simple hardware-friendly classifiers based on linear separators can achieve competitive performance while allowing real-time inference and fast training. Various solutions including batch and Online Sequential Learning are analyzed. Additionally, our experiments show that latency and performance can be further improved via dimensionality reduction and pre-training, respectively. The resulting system is implemented on two types of edge device, namely an edge accelerator and two smartphones. |
| first_indexed | 2024-10-01T07:34:13Z |
| format | Journal Article |
| id | ntu-10356/153915 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T07:34:13Z |
| publishDate | 2022 |
| record_format | dspace |
| spelling | ntu-10356/1539152022-06-03T05:18:19Z Emotion recognition on edge devices: training and deployment Pandelea, Vlad Ragusa, Edoardo Apicella, Tommaso Gastaldo, Paolo Cambria, Erik School of Computer Science and Engineering Engineering::Computer science and engineering Emotion Recognition Embedded Systems Emotion recognition, among other natural language processing tasks, has greatly benefited from the use of large transformer models. Deploying these models on resource-constrained devices, however, is a major challenge due to their computational cost. In this paper, we show that the combination of large transformers, as high-quality feature extractors, and simple hardware-friendly classifiers based on linear separators can achieve competitive performance while allowing real-time inference and fast training. Various solutions including batch and Online Sequential Learning are analyzed. Additionally, our experiments show that latency and performance can be further improved via dimensionality reduction and pre-training, respectively. The resulting system is implemented on two types of edge device, namely an edge accelerator and two smartphones. Agency for Science, Technology and Research (A*STAR) Published version This research is supported by the Agency for Science, Technology and Research (A*STAR) under its AME Programmatic Funding Scheme (Project #A18A2b0046). 2022-06-03T05:18:19Z 2022-06-03T05:18:19Z 2021 Journal Article Pandelea, V., Ragusa, E., Apicella, T., Gastaldo, P. & Cambria, E. (2021). Emotion recognition on edge devices: training and deployment. Sensors, 21(13), 4496-. https://dx.doi.org/10.3390/s21134496 1424-8220 https://hdl.handle.net/10356/153915 10.3390/s21134496 34209251 2-s2.0-85108880961 13 21 4496 en A18A2b0046 Sensors © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/). application/pdf |
| spellingShingle | Engineering::Computer science and engineering Emotion Recognition Embedded Systems Pandelea, Vlad Ragusa, Edoardo Apicella, Tommaso Gastaldo, Paolo Cambria, Erik Emotion recognition on edge devices: training and deployment |
| title | Emotion recognition on edge devices: training and deployment |
| title_full | Emotion recognition on edge devices: training and deployment |
| title_fullStr | Emotion recognition on edge devices: training and deployment |
| title_full_unstemmed | Emotion recognition on edge devices: training and deployment |
| title_short | Emotion recognition on edge devices: training and deployment |
| title_sort | emotion recognition on edge devices training and deployment |
| topic | Engineering::Computer science and engineering Emotion Recognition Embedded Systems |
| url | https://hdl.handle.net/10356/153915 |
| work_keys_str_mv | AT pandeleavlad emotionrecognitiononedgedevicestraininganddeployment AT ragusaedoardo emotionrecognitiononedgedevicestraininganddeployment AT apicellatommaso emotionrecognitiononedgedevicestraininganddeployment AT gastaldopaolo emotionrecognitiononedgedevicestraininganddeployment AT cambriaerik emotionrecognitiononedgedevicestraininganddeployment |