Vision encoders should be image size agnostic and task driven

TL;DR

This paper advocates for vision encoders that are adaptable to image size and task-specific, inspired by biological efficiency, and demonstrates initial feasibility through a proof-of-concept for image classification.

Contribution

Proposes a novel approach for vision encoders to be dynamic, size-agnostic, and task-driven, inspired by biological efficiency, with initial proof-of-concept results.

Findings

Feasibility of size-agnostic, task-driven vision encoders demonstrated

Initial proof-of-concept for image classification supports approach

Encourages development of more efficient, adaptive vision systems

Abstract

This position paper argues that the next generation of vision encoders should be image size agnostic and task driven. The source of our inspiration is biological. Not a structural aspect of biological vision, but a behavioral trait -- efficiency. We focus on a couple of ways in which vision in nature is efficient, but modern vision encoders not. We -- humans and animals -- deal with vast quantities of visual data, and need to be smart where we focus our limited energy -- it depends on the task. It is our belief that vision encoders should be dynamic and the computational complexity should depend on the task at hand rather than the size of the image. We, also, provide concrete first steps towards our vision -- a proof-of-concept solution for image classification. Despite classification being not very representative for what we are trying to achieve, it shows that our approach is feasible and promising.