Header

UZH-Logo

Maintenance Infos

Siamese Networks for Few-Shot Learning on Edge Embedded Devices


Lungu, Iulia Alexandra; Aimar, Alessandro; Hu, Yuhuang; Delbruck, Tobi; Liu, Shih-Chii (2020). Siamese Networks for Few-Shot Learning on Edge Embedded Devices. IEEE Journal on Emerging and Selected Topics in Circuits and Systems, 10(4):488-497.

Abstract

Edge artificial intelligence hardware targets mainly inference networks that have been pretrained on massive datasets. The field of few-shot learning looks for methods that allow a network to produce high accuracy even when only a few samples of each class are available. Siamese networks can be used to tackle few-shot learning problems and are unique because they do not require retraining on the new samples of the new classes. Therefore they are suitable for edge hardware accelerators which often do not include on-chip training capabilities. This work describes improvements to a baseline Siamese network and benchmarking of the improved network on edge platforms. The modifications to the baseline network included adding multi-resolution kernels, a hybrid training process as well a different embedding similarity computation method. This network shows an average accuracy improvement of up to 22% across 4 datasets in a 5-way, 1-shot classification task. Benchmarking results using three edge computing platforms (NVIDIA Jetson Nano, Coral Edge TPU and a custom convolutional neural network accelerator) show that a Siamese classifier can run on these devices at reasonable frame rates for real-time performance, between 3 frames per second (FPS) on Jetson Nano and 60 FPS on the Edge TPU. By increasing the weight sparsity during training, the inference time of a network with 25% weight sparsity increases by 10 FPS but with only 1% drop in accuracy.

Abstract

Edge artificial intelligence hardware targets mainly inference networks that have been pretrained on massive datasets. The field of few-shot learning looks for methods that allow a network to produce high accuracy even when only a few samples of each class are available. Siamese networks can be used to tackle few-shot learning problems and are unique because they do not require retraining on the new samples of the new classes. Therefore they are suitable for edge hardware accelerators which often do not include on-chip training capabilities. This work describes improvements to a baseline Siamese network and benchmarking of the improved network on edge platforms. The modifications to the baseline network included adding multi-resolution kernels, a hybrid training process as well a different embedding similarity computation method. This network shows an average accuracy improvement of up to 22% across 4 datasets in a 5-way, 1-shot classification task. Benchmarking results using three edge computing platforms (NVIDIA Jetson Nano, Coral Edge TPU and a custom convolutional neural network accelerator) show that a Siamese classifier can run on these devices at reasonable frame rates for real-time performance, between 3 frames per second (FPS) on Jetson Nano and 60 FPS on the Edge TPU. By increasing the weight sparsity during training, the inference time of a network with 25% weight sparsity increases by 10 FPS but with only 1% drop in accuracy.

Statistics

Citations

Altmetrics

Additional indexing

Item Type:Journal Article, refereed, original work
Communities & Collections:07 Faculty of Science > Institute of Neuroinformatics
Dewey Decimal Classification:570 Life sciences; biology
Scopus Subject Areas:Physical Sciences > Electrical and Electronic Engineering
Uncontrolled Keywords:Electrical and Electronic Engineering
Language:English
Date:1 December 2020
Deposited On:16 Feb 2021 08:49
Last Modified:17 Feb 2021 21:02
Publisher:Institute of Electrical and Electronics Engineers
ISSN:2156-3357
OA Status:Closed
Publisher DOI:https://doi.org/10.1109/jetcas.2020.3033155
Project Information:
  • : FunderSNSF
  • : Grant ID40B2-0_181010
  • : Project TitleVIPS: ultra-low power VIsual Perception System

Download

Full text not available from this repository.
View at publisher

Get full-text in a library