NCS based ultra low power optimized machine learning techniques for image classification

Tiwari, Naman; Mondal, Koushik

doi:10.1109/tensymp46218.2019.8971238

Cited by 6 publications

(4 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Naman et al, [20] optimized a model for classification of book cover images into genres and then compiled the model for inference on an NCS. Xing et al, [21] proposed a CNN with Raspberry Pi based system that runs a pre-trained inference model using NCS.…”

Section: Recent Work On Ncs For Deep Learning Inferencementioning

confidence: 99%

Implementation of a Modified U-Net for Medical Image Segmentation on Edge Devices

Ali

Shah

et al. 2022

IEEE Trans. Circuits Syst. II

View full text Add to dashboard Cite

Deep learning techniques, particularly convolutional neural networks, have shown great potential in computer vision and medical imaging applications. However, deep learning models are computationally demanding as they require enormous computational power and specialized processing hardware for model training. To make these models portable and compatible for prototyping, their implementation on low-power devices is imperative. In this work, we present the implementation of Modified U-Net on Intel/Movidius Neural Compute Stick 2 (NCS-2) for the segmentation of medical images. We selected U-Net because, in medical image segmentation, U-Net is a prominent model that provides improved performance for medical image segmentation even if the dataset size is small. The modified U-Net model is evaluated for performance in terms of dice score. Experiments are reported for segmentation task on three medical imaging datasets: BraTs dataset of brain MRI, heart MRI dataset, and Ziehl-Neelsen sputum smear microscopy image (ZNSDB) dataset. For the proposed model, we reduced the number of parameters from 30 million in the U-Net model to 0.49 million in the proposed architecture. Experimental results show that the modified U-Net provides comparable performance while requiring significantly lower resources and provides inference on the NCS-2. The maximum dice scores recorded are 0.96 for the BraTs dataset, 0.94 for the heart MRI dataset, and 0.74 for the ZNSDB dataset.

show abstract

Section: Recent Work On Ncs For Deep Learning Inferencementioning

confidence: 99%

Implementation of a Modified U-Net for Medical Image Segmentation on Edge Devices

Ali

Shah

et al. 2022

IEEE Trans. Circuits Syst. II

View full text Add to dashboard Cite

show abstract

“…These architectures include recurrent neural networks, stochastic networks, autoencoders etc. that are capable of handling complex radar images to low power edge computing devices for balanced and imbalanced datasets, as described in, [12][13][14]. Now with the advancements of Keras with the TensorFlow architecture, it is the beginning of another boom of IoT industries where data processing with the help of preprocessing quality management techniques offered a significant breakthrough as mentioned in [15].…”

Section: Review Of Literaturementioning

confidence: 99%

Different hybrid machine intelligence techniques for handling IoT‐based imbalanced data

Mohindru

Mondal

Banka

2021

CAAI Trans on Intel Tech

Self Cite

View full text Add to dashboard Cite

In the era of automatic task processing or designing complex algorithms, to analyse data, it is always pertinent to find real-life solutions using cutting-edge tools and techniques to generate insights into the data. The data-driven machine learning models are now offering more or less worthy results when they are certainly balanced in the input data sets. Imbalanced data occurs when an unequal distribution of classes occurs in the input datasets. Building a predictive model on the imbalanced data set would cause a model that appears to yield high accuracy but does not generalize well to the new data in the minority class. Now the time has come to look into the datasets which are not so-called 'balanced' in nature but such datasets are generally encountered frequently in a workspace. To prevent creating models with false levels of accuracy, the imbalanced data should be rearranged before creating a predictive model. Those data are, sometimes, voluminous, heterogeneous and complex in nature and generate from different autonomous sources with distributed and decentralized control. The driving force is to efficiently handle these data sets using latest tools and techniques for research and commercial insights. The present article provides different such tools and techniques, in different computing frameworks, to handle such Internet of Things and other related datasets to review common techniques for handling imbalanced data in data ecosystems and offers a comparative data modelling framework in Keras for balanced and imbalanced datasets.This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

show abstract

“…Moreover, there is still some progress in scaling down existing processors despite the slowing down of processor miniaturization because of impending physical limits (14). Examples of lightweight computing units include Google Coral's tensor processing unit (TPU) (15), Intel's Neural Compute Stick (16), and the JeVois smart camera, for example, used in (17). Until now, these processors have not yet brought visionbased navigation to very small robots, like lightweight nanocopters.…”

Section: Introductionmentioning

confidence: 99%

Visual route following for tiny autonomous robots

van Dijk,

De Wagter,

de Croon

2024

Sci. Robot.

View full text Add to dashboard Cite

Navigation is an essential capability for autonomous robots. In particular, visual navigation has been a major research topic in robotics because cameras are lightweight, power-efficient sensors that provide rich information on the environment. However, the main challenge of visual navigation is that it requires substantial computational power and memory for visual processing and storage of the results. As of yet, this has precluded its use on small, extremely resource-constrained robots such as lightweight drones. Inspired by the parsimony of natural intelligence, we propose an insect-inspired approach toward visual navigation that is specifically aimed at extremely resource-restricted robots. It is a route-following approach in which a robot’s outbound trajectory is stored as a collection of highly compressed panoramic images together with their spatial relationships as measured with odometry. During the inbound journey, the robot uses a combination of odometry and visual homing to return to the stored locations, with visual homing preventing the buildup of odometric drift. A main advancement of the proposed strategy is that the number of stored compressed images is minimized by spacing them apart as far as the accuracy of odometry allows. To demonstrate the suitability for small systems, we implemented the strategy on a tiny 56-gram drone. The drone could successfully follow routes up to 100 meters with a trajectory representation that consumed less than 20 bytes per meter. The presented method forms a substantial step toward the autonomous visual navigation of tiny robots, facilitating their more widespread application.

show abstract

NCS based ultra low power optimized machine learning techniques for image classification

Cited by 6 publications

References 6 publications

Implementation of a Modified U-Net for Medical Image Segmentation on Edge Devices

Implementation of a Modified U-Net for Medical Image Segmentation on Edge Devices

Different hybrid machine intelligence techniques for handling IoT‐based imbalanced data

Visual route following for tiny autonomous robots

Contact Info

Product

Resources

About