AutoCorrect: Deep Inductive Alignment of Noisy Geometric Annotations

Honglie Chen (University of Oxford), Weidi Xie (University of Oxford), Andrea Vedaldi (University of Oxford), Andrew Zisserman (University of Oxford)

Abstract
We propose AutoCorrect, a method to automatically learn object-annotation alignments from a dataset with annotations affected by geometric noise. The method is based on a consistency loss that enables deep neural networks to be trained, given only noisy annotations as input, to correct the annotations. When some noise-free annotations are available, we show that the consistency loss reduces to a stricter self-supervised loss. We also show that the method can implicitly leverage object symmetries to reduce the ambiguity arising in correcting noisy annotations. When multiple objects-annotation pairs are present in an image, we introduce a spatial memory map that allows the network to correct annotations sequentially, one at a time, while accounting for all other annotations in the image and corrections performed so far. Through ablation, we show the benefit of these contributions, demonstrating excellent result on geo-spatial imagery. Specifically, we show result using a new dataset of Railway tracks as well as the public Inria Building benchmarks, achieving new state-of-the-art results for the latter.

DOI
10.5244/C.33.96
https://dx.doi.org/10.5244/C.33.96

Files
Paper (PDF)
Supplementary material (PDF)

BibTeX