CaltechAUTHORS
  A Caltech Library Service

Synthetic Examples Improve Generalization for Rare Classes

Beery, Sara and Liu, Yang and Morris, Dan and Piavis, Jim and Kapoor, Ashish and Meister, Markus and Joshi, Neel and Perona, Pietro (2020) Synthetic Examples Improve Generalization for Rare Classes. In: 2020 IEEE Winter Conference on Applications of Computer Vision (WACV). IEEE , Piscataway, NJ, pp. 852-862. ISBN 978-1-7281-6553-0. https://resolver.caltech.edu/CaltechAUTHORS:20200526-140034764

[img] PDF - Submitted Version
See Usage Policy.

6Mb

Use this Persistent URL to link to this item: https://resolver.caltech.edu/CaltechAUTHORS:20200526-140034764

Abstract

The ability to detect and classify rare occurrences in images has important applications - for example, counting rare and endangered species when studying biodiversity, or detecting infrequent traffic scenarios that pose a danger to self-driving cars. Few-shot learning is an open problem: current computer vision systems struggle to categorize objects they have seen only rarely during training, and collecting a sufficient number of training examples of rare events is often challenging and expensive, and sometimes outright impossible. We explore in depth an approach to this problem: complementing the few available training images with ad-hoc simulated data.Our testbed is animal species classification, which has a real-world long-tailed distribution. We present two natural world simulators, and analyze the effect of different axes of variation in simulation, such as pose, lighting, model, and simulation method, and we prescribe best practices for efficiently incorporating simulated data for real-world performance gain. Our experiments reveal that synthetic data can considerably reduce error rates for classes that are rare, that as the amount of simulated data is increased, accuracy on the target class improves, and that high variation of simulated data provides maximum performance gain.


Item Type:Book Section
Related URLs:
URLURL TypeDescription
https://doi.org/10.1109/WACV45572.2020.9093570DOIArticle
https://arxiv.org/abs/1904.05916arXivDiscussion Paper
ORCID:
AuthorORCID
Beery, Sara0000-0002-2544-1844
Meister, Markus0000-0003-2136-6506
Perona, Pietro0000-0002-7583-5809
Additional Information:© 2020 IEEE. We would like to thank the USGS and NPS for providing data. This work was supported by NSFG RFP Grant No. 1745301, the views are those of the authors and do not necessarily reflect the views of the NSF. Compute provided by Microsoft AI for Earth and AWS.
Funders:
Funding AgencyGrant Number
NSF Graduate Research FellowshipDGE-1745301
Microsoft AI for EarthUNSPECIFIED
Amazon Web ServicesUNSPECIFIED
Record Number:CaltechAUTHORS:20200526-140034764
Persistent URL:https://resolver.caltech.edu/CaltechAUTHORS:20200526-140034764
Official Citation:S. Beery et al., "Synthetic Examples Improve Generalization for Rare Classes," 2020 IEEE Winter Conference on Applications of Computer Vision (WACV), Snowmass Village, CO, USA, 2020, pp. 852-862, doi: 10.1109/WACV45572.2020.9093570
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:103465
Collection:CaltechAUTHORS
Deposited By: Tony Diaz
Deposited On:26 May 2020 21:19
Last Modified:17 Nov 2020 17:44

Repository Staff Only: item control page