Year: 2018
Type: preprint
Abstract: We demonstrate the surprising strength of unimodal baselines in multimodal domains, and make concrete recommendations for best practices in future research. Where existing work often compares against random or majority class baselines, we argue that unimodal approaches better capture and reflect dataset biases and therefore provide an important comparison when assessing the performance of multimodal techniques. We present unimodal ablations on three recent datasets in visual navigation and QA, seeing an up to 29% absolute gain in performance over published baselines.
Source: arXiv (Cornell University)
Authors: Thomason, Jesse, Gordon, Daniel, Bisk, Yonatan
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Awards: III: Medium: Learning Multimodal Knowledge about Entities and Events, NRI: Collaborative Research: Experiential Learning for Robots: From Physics to Actions to Tasks, RI: Small: A Data-Driven Framework to Sketch-to-Text Generation, CAREER: Computation and Approximation in Structured Learning, CAREER: Active and Action-Centric Visual Understanding, G8066003835
OpenAlex ID: W4394652301
