Tampere University of Technology

TUTCRIS Research Portal

Adaptive Inference Using Hierarchical Convolutional Bag-of-Features for Low-Power Embedded Platforms

Research output: Chapter in Book/Report/Conference proceedingConference contributionScientificpeer-review

Standard

Adaptive Inference Using Hierarchical Convolutional Bag-of-Features for Low-Power Embedded Platforms. / Passalis, Nikolaos; Raitoharju, Jenni; Tefas, Anastasios; Gabbouj, Moncef.

2019 IEEE International Conference on Image Processing (ICIP). IEEE, 2019. p. 3048-3052 (IEEE International Conference on Image Processing).

Research output: Chapter in Book/Report/Conference proceedingConference contributionScientificpeer-review

Harvard

Passalis, N, Raitoharju, J, Tefas, A & Gabbouj, M 2019, Adaptive Inference Using Hierarchical Convolutional Bag-of-Features for Low-Power Embedded Platforms. in 2019 IEEE International Conference on Image Processing (ICIP). IEEE International Conference on Image Processing, IEEE, pp. 3048-3052, IEEE International Conference on Image Processing, 1/01/00. https://doi.org/10.1109/ICIP.2019.8803283

APA

Passalis, N., Raitoharju, J., Tefas, A., & Gabbouj, M. (2019). Adaptive Inference Using Hierarchical Convolutional Bag-of-Features for Low-Power Embedded Platforms. In 2019 IEEE International Conference on Image Processing (ICIP) (pp. 3048-3052). (IEEE International Conference on Image Processing). IEEE. https://doi.org/10.1109/ICIP.2019.8803283

Vancouver

Passalis N, Raitoharju J, Tefas A, Gabbouj M. Adaptive Inference Using Hierarchical Convolutional Bag-of-Features for Low-Power Embedded Platforms. In 2019 IEEE International Conference on Image Processing (ICIP). IEEE. 2019. p. 3048-3052. (IEEE International Conference on Image Processing). https://doi.org/10.1109/ICIP.2019.8803283

Author

Passalis, Nikolaos ; Raitoharju, Jenni ; Tefas, Anastasios ; Gabbouj, Moncef. / Adaptive Inference Using Hierarchical Convolutional Bag-of-Features for Low-Power Embedded Platforms. 2019 IEEE International Conference on Image Processing (ICIP). IEEE, 2019. pp. 3048-3052 (IEEE International Conference on Image Processing).

Bibtex - Download

@inproceedings{6b2b78d5e70346c28b90c9787d250468,
title = "Adaptive Inference Using Hierarchical Convolutional Bag-of-Features for Low-Power Embedded Platforms",
abstract = "Using early exits provide a straightforward way to implement models that can adapt on-the-fly to the available computational resources. However, early exits in many cases suffer from significant limitations, which often prohibit their practical application, especially when placed on convolutional layers with narrow receptive fields. In this work, we propose a method capable of overcoming these limitations by a) using a Bag-of-Features (BoF)-based pooling approach, that allows for keeping more information regarding the distribution of the extracted feature vectors, while also maintaining more spatial information and b) employing a simple, yet effective, hierarchical approach for designing the exits, allowing for efficiently re-using the information that was already extracted by the previous layers. It is experimentally demonstrated that the proposed approach leads to significant performance improvements, allowing early exits to be a more practical tool that can be used in many real-world embedded applications.",
keywords = "Feature extraction, Data mining, Convolutional codes, Adaptation models, Computational modeling, Convolution, Computer architecture, Early Exits, Adaptive Inference, Bag-of-Features, Lightweight Deep Learning",
author = "Nikolaos Passalis and Jenni Raitoharju and Anastasios Tefas and Moncef Gabbouj",
note = "EXT={"}Tefas, Anastasios{"}",
year = "2019",
month = "9",
doi = "10.1109/ICIP.2019.8803283",
language = "English",
isbn = "978-1-5386-6250-2",
series = "IEEE International Conference on Image Processing",
publisher = "IEEE",
pages = "3048--3052",
booktitle = "2019 IEEE International Conference on Image Processing (ICIP)",

}

RIS (suitable for import to EndNote) - Download

TY - GEN

T1 - Adaptive Inference Using Hierarchical Convolutional Bag-of-Features for Low-Power Embedded Platforms

AU - Passalis, Nikolaos

AU - Raitoharju, Jenni

AU - Tefas, Anastasios

AU - Gabbouj, Moncef

N1 - EXT="Tefas, Anastasios"

PY - 2019/9

Y1 - 2019/9

N2 - Using early exits provide a straightforward way to implement models that can adapt on-the-fly to the available computational resources. However, early exits in many cases suffer from significant limitations, which often prohibit their practical application, especially when placed on convolutional layers with narrow receptive fields. In this work, we propose a method capable of overcoming these limitations by a) using a Bag-of-Features (BoF)-based pooling approach, that allows for keeping more information regarding the distribution of the extracted feature vectors, while also maintaining more spatial information and b) employing a simple, yet effective, hierarchical approach for designing the exits, allowing for efficiently re-using the information that was already extracted by the previous layers. It is experimentally demonstrated that the proposed approach leads to significant performance improvements, allowing early exits to be a more practical tool that can be used in many real-world embedded applications.

AB - Using early exits provide a straightforward way to implement models that can adapt on-the-fly to the available computational resources. However, early exits in many cases suffer from significant limitations, which often prohibit their practical application, especially when placed on convolutional layers with narrow receptive fields. In this work, we propose a method capable of overcoming these limitations by a) using a Bag-of-Features (BoF)-based pooling approach, that allows for keeping more information regarding the distribution of the extracted feature vectors, while also maintaining more spatial information and b) employing a simple, yet effective, hierarchical approach for designing the exits, allowing for efficiently re-using the information that was already extracted by the previous layers. It is experimentally demonstrated that the proposed approach leads to significant performance improvements, allowing early exits to be a more practical tool that can be used in many real-world embedded applications.

KW - Feature extraction

KW - Data mining

KW - Convolutional codes

KW - Adaptation models

KW - Computational modeling

KW - Convolution

KW - Computer architecture

KW - Early Exits

KW - Adaptive Inference

KW - Bag-of-Features

KW - Lightweight Deep Learning

U2 - 10.1109/ICIP.2019.8803283

DO - 10.1109/ICIP.2019.8803283

M3 - Conference contribution

SN - 978-1-5386-6250-2

T3 - IEEE International Conference on Image Processing

SP - 3048

EP - 3052

BT - 2019 IEEE International Conference on Image Processing (ICIP)

PB - IEEE

ER -