ML Anthology

Pre-Gen Metrics: Predicting Caption Quality Metrics Without Generating Captions

Marc Tanti, Albert Gatt, Adrian Muscat
ECCVW 2018 pp. 114-123
doi:10.1007/978-3-030-11018-5_10 /eccvw/2018/tanti2018eccvw-pregen/

Abstract

Image caption generation systems are typically evaluated against reference outputs. We show that it is possible to predict output quality without generating the captions, based on the probability assigned by the neural model to the reference captions. Such pre-gen metrics are strongly correlated to standard evaluation metrics.

PDF ECCVW Semantic Scholar

Cite

Text

Tanti et al. "Pre-Gen Metrics: Predicting Caption Quality Metrics Without Generating Captions." European Conference on Computer Vision Workshops, 2018. doi:10.1007/978-3-030-11018-5_10

Markdown

[Tanti et al. "Pre-Gen Metrics: Predicting Caption Quality Metrics Without Generating Captions." European Conference on Computer Vision Workshops, 2018.](https://mlanthology.org/eccvw/2018/tanti2018eccvw-pregen/) doi:10.1007/978-3-030-11018-5_10

BibTeX

@inproceedings{tanti2018eccvw-pregen,
  title     = {{Pre-Gen Metrics: Predicting Caption Quality Metrics Without Generating Captions}},
  author    = {Tanti, Marc and Gatt, Albert and Muscat, Adrian},
  booktitle = {European Conference on Computer Vision Workshops},
  year      = {2018},
  pages     = {114-123},
  doi       = {10.1007/978-3-030-11018-5_10},
  url       = {https://mlanthology.org/eccvw/2018/tanti2018eccvw-pregen/}
}