Compressed Domain Image Classification Using a Dynamic-Rate Neural Network
Compressed domain image classification performs classification directly on compressive measurements acquired from the single-pixel camera, bypassing the image reconstruction step. It is of great importance for extending high-speed object detection and classification beyond the visible spectrum in a...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2020-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/9274326/ |
| _version_ | 1818917498474463232 |
|---|---|
| author | Yibo Xu Weidi Liu Kevin F. Kelly |
| author_facet | Yibo Xu Weidi Liu Kevin F. Kelly |
| author_sort | Yibo Xu |
| collection | DOAJ |
| description | Compressed domain image classification performs classification directly on compressive measurements acquired from the single-pixel camera, bypassing the image reconstruction step. It is of great importance for extending high-speed object detection and classification beyond the visible spectrum in a cost-effective manner especially for resource-limited platforms. Previous neural network methods require training a dedicated neural network for each different measurement rate (MR), which is costly in computation and storage. In this work, we develop an efficient training scheme that provides a neural network with dynamic-rate property, where a single neural network is capable of classifying over any MR within the range of interest with a given sensing matrix. This training scheme uses only a few selected MRs for training and the trained neural network is valid over the full range of MRs of interest. We demonstrate the performance of the dynamic-rate neural network on datasets of MNIST, CIFAR-10, Fashion-MNIST, COIL-100, and show that it generates approximately equal performance at each MR as that of a single-rate neural network valid only for one MR. Robustness to noise of the dynamic-rate model is also demonstrated. The dynamic-rate training scheme can be regarded as a general approach compatible with different types of sensing matrices, various neural network architectures, and is a valuable step towards wider adoption of compressive inference techniques and other compressive sensing related tasks via neural networks. |
| first_indexed | 2024-12-20T00:35:01Z |
| format | Article |
| id | doaj.art-0dc86d4898de481e9c5c0d994cffcb5f |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-12-20T00:35:01Z |
| publishDate | 2020-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-0dc86d4898de481e9c5c0d994cffcb5f2022-12-21T19:59:47ZengIEEEIEEE Access2169-35362020-01-01821771121772210.1109/ACCESS.2020.30418079274326Compressed Domain Image Classification Using a Dynamic-Rate Neural NetworkYibo Xu0https://orcid.org/0000-0002-3418-5387Weidi Liu1Kevin F. Kelly2Department of Electrical and Computer Engineering, Rice University, Houston, TX, USADepartment of Applied Physics, Rice University, Houston, TX, USADepartment of Electrical and Computer Engineering, Rice University, Houston, TX, USACompressed domain image classification performs classification directly on compressive measurements acquired from the single-pixel camera, bypassing the image reconstruction step. It is of great importance for extending high-speed object detection and classification beyond the visible spectrum in a cost-effective manner especially for resource-limited platforms. Previous neural network methods require training a dedicated neural network for each different measurement rate (MR), which is costly in computation and storage. In this work, we develop an efficient training scheme that provides a neural network with dynamic-rate property, where a single neural network is capable of classifying over any MR within the range of interest with a given sensing matrix. This training scheme uses only a few selected MRs for training and the trained neural network is valid over the full range of MRs of interest. We demonstrate the performance of the dynamic-rate neural network on datasets of MNIST, CIFAR-10, Fashion-MNIST, COIL-100, and show that it generates approximately equal performance at each MR as that of a single-rate neural network valid only for one MR. Robustness to noise of the dynamic-rate model is also demonstrated. The dynamic-rate training scheme can be regarded as a general approach compatible with different types of sensing matrices, various neural network architectures, and is a valuable step towards wider adoption of compressive inference techniques and other compressive sensing related tasks via neural networks.https://ieeexplore.ieee.org/document/9274326/Compressive sensingimage classificationsingle-pixel cameraneural networks |
| spellingShingle | Yibo Xu Weidi Liu Kevin F. Kelly Compressed Domain Image Classification Using a Dynamic-Rate Neural Network IEEE Access Compressive sensing image classification single-pixel camera neural networks |
| title | Compressed Domain Image Classification Using a Dynamic-Rate Neural Network |
| title_full | Compressed Domain Image Classification Using a Dynamic-Rate Neural Network |
| title_fullStr | Compressed Domain Image Classification Using a Dynamic-Rate Neural Network |
| title_full_unstemmed | Compressed Domain Image Classification Using a Dynamic-Rate Neural Network |
| title_short | Compressed Domain Image Classification Using a Dynamic-Rate Neural Network |
| title_sort | compressed domain image classification using a dynamic rate neural network |
| topic | Compressive sensing image classification single-pixel camera neural networks |
| url | https://ieeexplore.ieee.org/document/9274326/ |
| work_keys_str_mv | AT yiboxu compresseddomainimageclassificationusingadynamicrateneuralnetwork AT weidiliu compresseddomainimageclassificationusingadynamicrateneuralnetwork AT kevinfkelly compresseddomainimageclassificationusingadynamicrateneuralnetwork |