NoFADE: Analyzing Diminishing Returns on CO2 Investment
This addresses the issue of inefficient CO2 investment in CV research for the computer vision community, though it is incremental as it builds on existing analysis of performance saturation.
The paper tackles the problem of diminishing returns in computer vision (CV) methods by analyzing their saturation across tasks and proposes NoFADE, an entropy-based metric to quantify model-dataset-complexity relationships, showing that some tasks are reaching or are almost fully saturated.
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.