Explaining the Impact of Training on Vision Models via Activation Clustering

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Boubekki, Ahcène, Fadel, Samuel G., Mair, Sebastian
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913939449708544
author Boubekki, Ahcène
Fadel, Samuel G.
Mair, Sebastian
author_facet Boubekki, Ahcène
Fadel, Samuel G.
Mair, Sebastian
contents This paper introduces Neuro-Activated Vision Explanations (NAVE), a method for extracting and visualizing the internal representations of vision model encoders. By clustering feature activations, NAVE provides insights into learned semantics without fine-tuning. Using object localization, we show that NAVE's concepts align with image semantics. Through extensive experiments, we analyze the impact of training strategies and architectures on encoder representation capabilities. Additionally, we apply NAVE to study training artifacts in vision transformers and reveal how weak training strategies and spurious correlations degrade model performance. Our findings establish NAVE as a valuable tool for post-hoc model inspection and improving transparency in vision models.
format Preprint
id arxiv_https___arxiv_org_abs_2411_19700
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Explaining the Impact of Training on Vision Models via Activation Clustering
Boubekki, Ahcène
Fadel, Samuel G.
Mair, Sebastian
Computer Vision and Pattern Recognition
Machine Learning
This paper introduces Neuro-Activated Vision Explanations (NAVE), a method for extracting and visualizing the internal representations of vision model encoders. By clustering feature activations, NAVE provides insights into learned semantics without fine-tuning. Using object localization, we show that NAVE's concepts align with image semantics. Through extensive experiments, we analyze the impact of training strategies and architectures on encoder representation capabilities. Additionally, we apply NAVE to study training artifacts in vision transformers and reveal how weak training strategies and spurious correlations degrade model performance. Our findings establish NAVE as a valuable tool for post-hoc model inspection and improving transparency in vision models.
title Explaining the Impact of Training on Vision Models via Activation Clustering
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2411.19700