Quantitative Evaluation of White & Black Box Interpretability Methods for Image Classification

The field of interpretability in Deep Learning faces significant challenges due to the lack of standard metrics for systematically evaluating and comparing interpretability methods. The absence of quantifiable measures impedes practitioners ability to select the most suitable methods and models for...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Stanchi, Oscar Agustín, Ronchetti, Franco, Dal Bianco, Pedro Alejandro, Ríos, Gastón Gustavo, Hasperué, Waldo, Puig Valls, Domenec, Rashwan, Hatem, Quiroga, Facundo Manuel
Formato:	Objeto de conferencia
Lenguaje:	Inglés
Publicado:	2024
Materias:	Ciencias Informáticas Ablation Black Box Computer Vision Deep Learning Interpretability Quantitative Measure White Box
Acceso en línea:	http://sedici.unlp.edu.ar/handle/10915/176288
Aporte de:	SEDICI (UNLP) de Universidad Nacional de La Plata

id	I19-R120-10915-176288
record_format	dspace
spelling	I19-R120-10915-1762882025-02-07T20:04:55Z http://sedici.unlp.edu.ar/handle/10915/176288 Quantitative Evaluation of White & Black Box Interpretability Methods for Image Classification Stanchi, Oscar Agustín Ronchetti, Franco Dal Bianco, Pedro Alejandro Ríos, Gastón Gustavo Hasperué, Waldo Puig Valls, Domenec Rashwan, Hatem Quiroga, Facundo Manuel 2024-10 2024 2025-02-07T17:25:08Z en Ciencias Informáticas Ablation Black Box Computer Vision Deep Learning Interpretability Quantitative Measure White Box The field of interpretability in Deep Learning faces significant challenges due to the lack of standard metrics for systematically evaluating and comparing interpretability methods. The absence of quantifiable measures impedes practitioners ability to select the most suitable methods and models for their specific tasks. To address this issue, we propose the Pixel Erosion and Dilation Score, a novel metric designed to assess the robustness of model explanations. Our approach involves applying iterative erosion and dilation processes to heatmaps generated by various interpretability methods, thereby using them to hide and show the important regions of a image to the network, allowing for a coherent and interpretable evaluation of model decision-making processes. We conduct quantitative ablation tests using our metric on the ImageNet dataset with both VGG16 and ResNet18 models. The results reveal that our new measure provides a numerical and intuitive means for comparing interpretability methods and models, facilitating more informed decision-making for practitioner. Red de Universidades con Carreras en Informática Objeto de conferencia Objeto de conferencia http://creativecommons.org/licenses/by-nc-sa/4.0/ Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) application/pdf 125-134
institution	Universidad Nacional de La Plata
institution_str	I-19
repository_str	R-120
collection	SEDICI (UNLP)
language	Inglés
topic	Ciencias Informáticas Ablation Black Box Computer Vision Deep Learning Interpretability Quantitative Measure White Box
spellingShingle	Ciencias Informáticas Ablation Black Box Computer Vision Deep Learning Interpretability Quantitative Measure White Box Stanchi, Oscar Agustín Ronchetti, Franco Dal Bianco, Pedro Alejandro Ríos, Gastón Gustavo Hasperué, Waldo Puig Valls, Domenec Rashwan, Hatem Quiroga, Facundo Manuel Quantitative Evaluation of White & Black Box Interpretability Methods for Image Classification
topic_facet	Ciencias Informáticas Ablation Black Box Computer Vision Deep Learning Interpretability Quantitative Measure White Box
description	The field of interpretability in Deep Learning faces significant challenges due to the lack of standard metrics for systematically evaluating and comparing interpretability methods. The absence of quantifiable measures impedes practitioners ability to select the most suitable methods and models for their specific tasks. To address this issue, we propose the Pixel Erosion and Dilation Score, a novel metric designed to assess the robustness of model explanations. Our approach involves applying iterative erosion and dilation processes to heatmaps generated by various interpretability methods, thereby using them to hide and show the important regions of a image to the network, allowing for a coherent and interpretable evaluation of model decision-making processes. We conduct quantitative ablation tests using our metric on the ImageNet dataset with both VGG16 and ResNet18 models. The results reveal that our new measure provides a numerical and intuitive means for comparing interpretability methods and models, facilitating more informed decision-making for practitioner.
format	Objeto de conferencia Objeto de conferencia
author	Stanchi, Oscar Agustín Ronchetti, Franco Dal Bianco, Pedro Alejandro Ríos, Gastón Gustavo Hasperué, Waldo Puig Valls, Domenec Rashwan, Hatem Quiroga, Facundo Manuel
author_facet	Stanchi, Oscar Agustín Ronchetti, Franco Dal Bianco, Pedro Alejandro Ríos, Gastón Gustavo Hasperué, Waldo Puig Valls, Domenec Rashwan, Hatem Quiroga, Facundo Manuel
author_sort	Stanchi, Oscar Agustín
title	Quantitative Evaluation of White & Black Box Interpretability Methods for Image Classification
title_short	Quantitative Evaluation of White & Black Box Interpretability Methods for Image Classification
title_full	Quantitative Evaluation of White & Black Box Interpretability Methods for Image Classification
title_fullStr	Quantitative Evaluation of White & Black Box Interpretability Methods for Image Classification
title_full_unstemmed	Quantitative Evaluation of White & Black Box Interpretability Methods for Image Classification
title_sort	quantitative evaluation of white & black box interpretability methods for image classification
publishDate	2024
url	http://sedici.unlp.edu.ar/handle/10915/176288
work_keys_str_mv	AT stanchioscaragustin quantitativeevaluationofwhiteblackboxinterpretabilitymethodsforimageclassification AT ronchettifranco quantitativeevaluationofwhiteblackboxinterpretabilitymethodsforimageclassification AT dalbiancopedroalejandro quantitativeevaluationofwhiteblackboxinterpretabilitymethodsforimageclassification AT riosgastongustavo quantitativeevaluationofwhiteblackboxinterpretabilitymethodsforimageclassification AT hasperuewaldo quantitativeevaluationofwhiteblackboxinterpretabilitymethodsforimageclassification AT puigvallsdomenec quantitativeevaluationofwhiteblackboxinterpretabilitymethodsforimageclassification AT rashwanhatem quantitativeevaluationofwhiteblackboxinterpretabilitymethodsforimageclassification AT quirogafacundomanuel quantitativeevaluationofwhiteblackboxinterpretabilitymethodsforimageclassification
_version_	1845116778679631872

Quantitative Evaluation of White & Black Box Interpretability Methods for Image Classification

Ejemplares similares