Skip navigation
Please use this identifier to cite or link to this item: http://repository.iitr.ac.in/handle/123456789/16685
Title: Minimalistic Image Signal Processing for Deep Learning Applications
Authors: Lubana E.S.
DIck R.P.
Aggarwal V.
Pradhan, Pyari Mohan
Published in: Proceedings of International Conference on Image Processing, ICIP
Abstract: In-sensor energy-efficient deep learning accelerators have the potential to enable the use of deep neural networks in embedded vision applications. However, their negative impact on accuracy has been severely underestimated. The inference pipeline used in prior in-sensor deep learning accelerators bypasses the image signal processor (ISP), thereby disrupting the conventional vision pipeline and undermining accuracy of machine learning algorithms trained on conventional, post-ISP datasets. For example, the detection accuracy of an off-the-shelf Faster RCNN algorithm in a vehicle detection scenario reduces by 60%. To make in-sensor accelerators practical, we describe energy-efficient operations that yield most of the benefits of an ISP and reduce covariate shift between the training (ISP processed images) and target (RAW images) distributions. For the vehicle detection problem, our approach improves accuracy by 25-60%. Relative to the conventional ISP pipeline, energy consumption and response time improve by 30% and 34%, respectively. © 2019 IEEE.
Citation: Proceedings of International Conference on Image Processing, ICIP, (2019), 4165- 4169
URI: https://doi.org/10.1109/ICIP.2019.8803645
http://repository.iitr.ac.in/handle/123456789/16685
Issue Date: 2019
Publisher: IEEE Computer Society
Keywords: Covariate shift
Deep learning accelerators
Image signal processor
RAW images
ISBN: 9.78154E+12
ISSN: 15224880
Author Scopus IDs: 57202045941
7202246612
56861825500
26639724100
Author Affiliations: Lubana, E.S., University of Michigan, Ann Arbor, United States, Indian Institute of Technology, Roorkee, India
DIck, R.P., University of Michigan, Ann Arbor, United States
Aggarwal, V., Indian Institute of Technology, Roorkee, India
Pradhan, P.M., Indian Institute of Technology, Roorkee, India
Appears in Collections:Conference Publications [ECE]

Files in This Item:
There are no files associated with this item.
Show full item record


Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.