Abstract
Compressive imagers, e.g. the single-pixel camera (SPC), acquire measurements in the form of random projections of the scene instead of pixel intensities. Compressive Sensing (CS) theory allows accurate reconstruction of the image even from a small number of such projections. However, in practice, most reconstruction algorithms perform poorly at low measurement rates and are computationally very expensive. But perfect reconstruction is not the goal of high-level computer vision applications. Instead, we are interested in only determining certain properties of the image. Recent work has shown that effective inference is possible directly from the compressive measurements, without reconstruction, using correlational features. In this paper, we show that convolutional neural networks (CNNs) can be employed to extract discriminative non-linear features directly from CS measurements. Using these features, we demonstrate that effective high-level inference can be performed. Experimentally, using hand written digit recognition (MNIST dataset) and image recognition (ImageNet) as examples, we show that recognition is possible even at low measurement rates of about 0.1.
| Original language | English (US) |
|---|---|
| Title of host publication | 2016 IEEE International Conference on Image Processing, ICIP 2016 - Proceedings |
| Publisher | IEEE Computer Society |
| Pages | 1913-1917 |
| Number of pages | 5 |
| Volume | 2016-August |
| ISBN (Electronic) | 9781467399616 |
| DOIs | |
| State | Published - Aug 3 2016 |
| Event | 23rd IEEE International Conference on Image Processing, ICIP 2016 - Phoenix, United States Duration: Sep 25 2016 → Sep 28 2016 |
Other
| Other | 23rd IEEE International Conference on Image Processing, ICIP 2016 |
|---|---|
| Country | United States |
| City | Phoenix |
| Period | 9/25/16 → 9/28/16 |
Fingerprint
Keywords
- Compressive sensing
- Neural networks
ASJC Scopus subject areas
- Software
- Computer Vision and Pattern Recognition
- Signal Processing
Cite this
Direct inference on compressive measurements using convolutional neural networks. / Lohit, Suhas; Kulkarni, Kuldeep; Turaga, Pavan.
2016 IEEE International Conference on Image Processing, ICIP 2016 - Proceedings. Vol. 2016-August IEEE Computer Society, 2016. p. 1913-1917 7532691.Research output: Chapter in Book/Report/Conference proceeding › Conference contribution
}
TY - GEN
T1 - Direct inference on compressive measurements using convolutional neural networks
AU - Lohit, Suhas
AU - Kulkarni, Kuldeep
AU - Turaga, Pavan
PY - 2016/8/3
Y1 - 2016/8/3
N2 - Compressive imagers, e.g. the single-pixel camera (SPC), acquire measurements in the form of random projections of the scene instead of pixel intensities. Compressive Sensing (CS) theory allows accurate reconstruction of the image even from a small number of such projections. However, in practice, most reconstruction algorithms perform poorly at low measurement rates and are computationally very expensive. But perfect reconstruction is not the goal of high-level computer vision applications. Instead, we are interested in only determining certain properties of the image. Recent work has shown that effective inference is possible directly from the compressive measurements, without reconstruction, using correlational features. In this paper, we show that convolutional neural networks (CNNs) can be employed to extract discriminative non-linear features directly from CS measurements. Using these features, we demonstrate that effective high-level inference can be performed. Experimentally, using hand written digit recognition (MNIST dataset) and image recognition (ImageNet) as examples, we show that recognition is possible even at low measurement rates of about 0.1.
AB - Compressive imagers, e.g. the single-pixel camera (SPC), acquire measurements in the form of random projections of the scene instead of pixel intensities. Compressive Sensing (CS) theory allows accurate reconstruction of the image even from a small number of such projections. However, in practice, most reconstruction algorithms perform poorly at low measurement rates and are computationally very expensive. But perfect reconstruction is not the goal of high-level computer vision applications. Instead, we are interested in only determining certain properties of the image. Recent work has shown that effective inference is possible directly from the compressive measurements, without reconstruction, using correlational features. In this paper, we show that convolutional neural networks (CNNs) can be employed to extract discriminative non-linear features directly from CS measurements. Using these features, we demonstrate that effective high-level inference can be performed. Experimentally, using hand written digit recognition (MNIST dataset) and image recognition (ImageNet) as examples, we show that recognition is possible even at low measurement rates of about 0.1.
KW - Compressive sensing
KW - Neural networks
UR - http://www.scopus.com/inward/record.url?scp=85006837039&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85006837039&partnerID=8YFLogxK
U2 - 10.1109/ICIP.2016.7532691
DO - 10.1109/ICIP.2016.7532691
M3 - Conference contribution
VL - 2016-August
SP - 1913
EP - 1917
BT - 2016 IEEE International Conference on Image Processing, ICIP 2016 - Proceedings
PB - IEEE Computer Society
ER -