ML Anthology

Dealing with Small Data and Training Blind Spots in the Manhattan World

Wajahat Hussain, Javier Civera, Luis Montano, Martial Hebert
WACV 2016 pp. 1-9
doi:10.1109/WACV.2016.7477649 /wacv/2016/hussain2016wacv-dealing/

Abstract

Leveraging Manhattan assumption we generate metrically rectified novel views from a single image, even for non-box scenarios. Our novel views enable the already trained classifiers to handle training data missing views (blind spots) without additional training. We demonstrate this on end-to-end scene text spotting under perspective. Additionally, utilizing our fronto-parallel views, we discover unsuspended invariant mid-level patches given a few widely separated training examples (small data domain). These invariant patches outperform various baselines on small data image retrieval challenge.

WACV Semantic Scholar

Cite

Text

Hussain et al. "Dealing with Small Data and Training Blind Spots in the Manhattan World." IEEE/CVF Winter Conference on Applications of Computer Vision, 2016. doi:10.1109/WACV.2016.7477649

Markdown

[Hussain et al. "Dealing with Small Data and Training Blind Spots in the Manhattan World." IEEE/CVF Winter Conference on Applications of Computer Vision, 2016.](https://mlanthology.org/wacv/2016/hussain2016wacv-dealing/) doi:10.1109/WACV.2016.7477649

BibTeX

@inproceedings{hussain2016wacv-dealing,
  title     = {{Dealing with Small Data and Training Blind Spots in the Manhattan World}},
  author    = {Hussain, Wajahat and Civera, Javier and Montano, Luis and Hebert, Martial},
  booktitle = {IEEE/CVF Winter Conference on Applications of Computer Vision},
  year      = {2016},
  pages     = {1-9},
  doi       = {10.1109/WACV.2016.7477649},
  url       = {https://mlanthology.org/wacv/2016/hussain2016wacv-dealing/}
}