NoFADE: Analyzing Diminishing Returns on CO2 Investment

TL;DR

This paper introduces NoFADE, a new entropy-based metric to analyze diminishing returns in computer vision, revealing saturation levels of tasks and enabling fair model-dataset comparisons to reduce environmental impact.

Contribution

The paper proposes NoFADE, a novel metric for quantifying model-dataset complexity relationships and assessing saturation in CV tasks.

Findings

Some CV tasks are reaching saturation.

Other tasks are nearly fully saturated.

NoFADE provides a platform for fair model and dataset comparison.

Abstract

Climate change continues to be a pressing issue that currently affects society at-large. It is important that we as a society, including the Computer Vision (CV) community take steps to limit our impact on the environment. In this paper, we (a) analyze the effect of diminishing returns on CV methods, and (b) propose a \textit{``NoFADE''}: a novel entropy-based metric to quantify model--dataset--complexity relationships. We show that some CV tasks are reaching saturation, while others are almost fully saturated. In this light, NoFADE allows the CV community to compare models and datasets on a similar basis, establishing an agnostic platform.