Similarity of Processing Steps in Vision Model Representations

TL;DR

This paper investigates how different vision models develop similar representations, focusing on their processing steps and the evolution of their internal representations across layers.

Contribution

It introduces a method to analyze and compare the processing steps leading to convergent representations in vision models, highlighting differences between model types.

Findings

Layers at similar positions have the most similar representations.

Classifier models discard low-level image information in final layers.

Transformers change representations more smoothly across layers.

Abstract

Recent literature suggests that the bigger the model, the more likely it is to converge to similar, ``universal'' representations, despite different training objectives, datasets, or modalities. While this literature shows that there is an area where model representations are similar, we study here how vision models might get to those representations -- in particular, do they also converge to the same intermediate steps and operations? We therefore study the processes that lead to convergent representations in different models. First, we quantify distance between different model representations at different stages. We follow the evolution of distances between models throughout processing, identifying the processing steps which are most different between models. We find that while layers at similar positions in different models have the most similar representations, strong differences remain. Classifier models, unlike the others, will discard information about low-level image statistics in their final layers. CNN- and transformer-based models also behave differently, with transformer models applying smoother changes to representations from one layer to the next. These distinctions clarify the level and nature of convergence between model representations, and enables a more qualitative account of the underlying processes in image models.