Discriminate-and-Rectify Encoders: Learning from Image Transformation Sets
Name
CBMM-Memo-062.pdf
Size
9.37 MB
Format
Adobe PDF
Checksum (MD5)
e79a9451828a43e2941a9b2ca7d3fc36
Author(s)
Tachetti, Andrea
Voinea, Stephen
Evangelopoulos, Georgios
Date Issued
March 13, 2017
Publisher
Center for Brains, Minds and Machines (CBMM), arXiv
Citation
arXiv:1703.04775v1
Series/Report no.
CBMM Memo Series;062
Abstract
The complexity of a learning task is increased by transformations in the input space that preserve class identity. Visual object recognition for example is affected by changes in viewpoint, scale, illumination or planar transformations. While drastically altering the visual appearance, these changes are orthogonal to recognition and should not be reflected in the representation or feature encoding used for learning. We introduce a framework for weakly supervised learning of image embeddings that are robust to transformations and selective to the class distribution, using sets of transforming examples (orbit sets), deep parametrizations and a novel orbit-based loss. The proposed loss combines a discriminative, contrastive part for orbits with a reconstruction error that learns to rectify orbit transformations. The learned embeddings are evaluated in distance metric-based tasks, such as one-shot classification under geometric transformations, as well as face verification and retrieval under more realistic visual variability. Our results suggest that orbit sets, suitably computed or observed, can be used for efficient, weakly-supervised learning of semantically relevant image embeddings.
Subjects
supervised learning
object recognition
machine learning
Terms of Use
Attribution-NonCommercial-ShareAlike 3.0 United States
Persistent DSpace Link
