Loading…
Multi-Level XAI-Driven MLOps Pipeline for the Adjustment of Fruit and Vegetable Classifiers
In this paper, we present a machine learning operations (MLOps) pipeline that exploits explainable artificial intelligence (XAI) to adjust deep neural network (DNN)-based fruit and vegetable image classifiers to data observed at super-market self-checkouts. DNNs are currently the most successful AI...
Saved in:
Main Authors: | , , , , , |
---|---|
Format: | Conference Proceeding |
Language: | English |
Subjects: | |
Online Access: | Request full text |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
cited_by | |
---|---|
cites | |
container_end_page | 6 |
container_issue | |
container_start_page | 1 |
container_title | |
container_volume | |
creator | Iriarte, Francisco J. Ortiz, Miguel E. Unzueta, Luis Martinez, Javier Zaldivar, Javier Arganda-Carreras, Ignacio |
description | In this paper, we present a machine learning operations (MLOps) pipeline that exploits explainable artificial intelligence (XAI) to adjust deep neural network (DNN)-based fruit and vegetable image classifiers to data observed at super-market self-checkouts. DNNs are currently the most successful AI models for several automation tasks, including fruit and vegetable classification. However, adjusting them to on-site observations to avoid model drift is challenging, as they work as black boxes that take inputs and return results without showing the processes that lead to their responses. State-of-the-art XAI techniques could help mitigate this problem, but integrating them in MLOps pipelines is also challenging due to their high computational cost. We propose multi-task B-cos (MTBC) DNNs integrated at different pipeline levels to obtain automated information about the system's on-site performance. MTBC DNNs allow indentification of fruit and vegetable types with contextual attributes in real time and generation of contribution maps that highlight task-relevant features for enhanced model adjustment. Experimental results show that MTBC DNNs obtain similar accuracy and performance to the equivalent baseline DNNs for fruit and vegetable classification but with the added benefit of explainability, supporting MLOps processes such as drift detection and exploratory data analysis. |
doi_str_mv | 10.1109/IS61756.2024.10705202 |
format | conference_proceeding |
fullrecord | <record><control><sourceid>ieee_CHZPO</sourceid><recordid>TN_cdi_ieee_primary_10705202</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><ieee_id>10705202</ieee_id><sourcerecordid>10705202</sourcerecordid><originalsourceid>FETCH-ieee_primary_107052023</originalsourceid><addsrcrecordid>eNqFjs1qwkAURqeFQqXNG1S4L5B4Z2IymaX4g4JSwVIKXciIN3pljGFmIvTt68KuCx-cA2fzCdGXmEmJZrDYlFIXZaZQDTOJGoubPYjEaFPlBd5mKvUoekqXOjUVqmeRhHBCxFzhUCrZE9-rzkVOl3QlB1-jRTrxfKUGVsv3NsCaW3LcENQXD_FIMNqfuhDP1ES41DDzHUewzR4-6UDR7hzB2NkQuGby4VU81dYFSu58EW-z6cd4njIRbVvPZ-t_tn-_83_yL73_RRM</addsrcrecordid><sourcetype>Publisher</sourcetype><iscdi>true</iscdi><recordtype>conference_proceeding</recordtype></control><display><type>conference_proceeding</type><title>Multi-Level XAI-Driven MLOps Pipeline for the Adjustment of Fruit and Vegetable Classifiers</title><source>IEEE Xplore All Conference Series</source><creator>Iriarte, Francisco J. ; Ortiz, Miguel E. ; Unzueta, Luis ; Martinez, Javier ; Zaldivar, Javier ; Arganda-Carreras, Ignacio</creator><creatorcontrib>Iriarte, Francisco J. ; Ortiz, Miguel E. ; Unzueta, Luis ; Martinez, Javier ; Zaldivar, Javier ; Arganda-Carreras, Ignacio</creatorcontrib><description>In this paper, we present a machine learning operations (MLOps) pipeline that exploits explainable artificial intelligence (XAI) to adjust deep neural network (DNN)-based fruit and vegetable image classifiers to data observed at super-market self-checkouts. DNNs are currently the most successful AI models for several automation tasks, including fruit and vegetable classification. However, adjusting them to on-site observations to avoid model drift is challenging, as they work as black boxes that take inputs and return results without showing the processes that lead to their responses. State-of-the-art XAI techniques could help mitigate this problem, but integrating them in MLOps pipelines is also challenging due to their high computational cost. We propose multi-task B-cos (MTBC) DNNs integrated at different pipeline levels to obtain automated information about the system's on-site performance. MTBC DNNs allow indentification of fruit and vegetable types with contextual attributes in real time and generation of contribution maps that highlight task-relevant features for enhanced model adjustment. Experimental results show that MTBC DNNs obtain similar accuracy and performance to the equivalent baseline DNNs for fruit and vegetable classification but with the added benefit of explainability, supporting MLOps processes such as drift detection and exploratory data analysis.</description><identifier>EISSN: 2767-9802</identifier><identifier>EISBN: 9798350350982</identifier><identifier>DOI: 10.1109/IS61756.2024.10705202</identifier><language>eng</language><publisher>IEEE</publisher><subject>Adaptation models ; Artificial neural networks ; Computational modeling ; Computer Vision ; Explainable AI ; Fruit and Vegetable Classification ; MLOps ; Object detection ; Pipelines ; Production ; Real-time systems ; Robustness ; Scalability ; XAI</subject><ispartof>2024 IEEE 12th International Conference on Intelligent Systems (IS), 2024, p.1-6</ispartof><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://ieeexplore.ieee.org/document/10705202$$EHTML$$P50$$Gieee$$H</linktohtml><link.rule.ids>309,310,780,784,789,790,27925,54555,54932</link.rule.ids><linktorsrc>$$Uhttps://ieeexplore.ieee.org/document/10705202$$EView_record_in_IEEE$$FView_record_in_$$GIEEE</linktorsrc></links><search><creatorcontrib>Iriarte, Francisco J.</creatorcontrib><creatorcontrib>Ortiz, Miguel E.</creatorcontrib><creatorcontrib>Unzueta, Luis</creatorcontrib><creatorcontrib>Martinez, Javier</creatorcontrib><creatorcontrib>Zaldivar, Javier</creatorcontrib><creatorcontrib>Arganda-Carreras, Ignacio</creatorcontrib><title>Multi-Level XAI-Driven MLOps Pipeline for the Adjustment of Fruit and Vegetable Classifiers</title><title>2024 IEEE 12th International Conference on Intelligent Systems (IS)</title><addtitle>IS</addtitle><description>In this paper, we present a machine learning operations (MLOps) pipeline that exploits explainable artificial intelligence (XAI) to adjust deep neural network (DNN)-based fruit and vegetable image classifiers to data observed at super-market self-checkouts. DNNs are currently the most successful AI models for several automation tasks, including fruit and vegetable classification. However, adjusting them to on-site observations to avoid model drift is challenging, as they work as black boxes that take inputs and return results without showing the processes that lead to their responses. State-of-the-art XAI techniques could help mitigate this problem, but integrating them in MLOps pipelines is also challenging due to their high computational cost. We propose multi-task B-cos (MTBC) DNNs integrated at different pipeline levels to obtain automated information about the system's on-site performance. MTBC DNNs allow indentification of fruit and vegetable types with contextual attributes in real time and generation of contribution maps that highlight task-relevant features for enhanced model adjustment. Experimental results show that MTBC DNNs obtain similar accuracy and performance to the equivalent baseline DNNs for fruit and vegetable classification but with the added benefit of explainability, supporting MLOps processes such as drift detection and exploratory data analysis.</description><subject>Adaptation models</subject><subject>Artificial neural networks</subject><subject>Computational modeling</subject><subject>Computer Vision</subject><subject>Explainable AI</subject><subject>Fruit and Vegetable Classification</subject><subject>MLOps</subject><subject>Object detection</subject><subject>Pipelines</subject><subject>Production</subject><subject>Real-time systems</subject><subject>Robustness</subject><subject>Scalability</subject><subject>XAI</subject><issn>2767-9802</issn><isbn>9798350350982</isbn><fulltext>true</fulltext><rsrctype>conference_proceeding</rsrctype><creationdate>2024</creationdate><recordtype>conference_proceeding</recordtype><sourceid>6IE</sourceid><recordid>eNqFjs1qwkAURqeFQqXNG1S4L5B4Z2IymaX4g4JSwVIKXciIN3pljGFmIvTt68KuCx-cA2fzCdGXmEmJZrDYlFIXZaZQDTOJGoubPYjEaFPlBd5mKvUoekqXOjUVqmeRhHBCxFzhUCrZE9-rzkVOl3QlB1-jRTrxfKUGVsv3NsCaW3LcENQXD_FIMNqfuhDP1ES41DDzHUewzR4-6UDR7hzB2NkQuGby4VU81dYFSu58EW-z6cd4njIRbVvPZ-t_tn-_83_yL73_RRM</recordid><startdate>20240829</startdate><enddate>20240829</enddate><creator>Iriarte, Francisco J.</creator><creator>Ortiz, Miguel E.</creator><creator>Unzueta, Luis</creator><creator>Martinez, Javier</creator><creator>Zaldivar, Javier</creator><creator>Arganda-Carreras, Ignacio</creator><general>IEEE</general><scope>6IE</scope><scope>6IL</scope><scope>CBEJK</scope><scope>RIE</scope><scope>RIL</scope></search><sort><creationdate>20240829</creationdate><title>Multi-Level XAI-Driven MLOps Pipeline for the Adjustment of Fruit and Vegetable Classifiers</title><author>Iriarte, Francisco J. ; Ortiz, Miguel E. ; Unzueta, Luis ; Martinez, Javier ; Zaldivar, Javier ; Arganda-Carreras, Ignacio</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-ieee_primary_107052023</frbrgroupid><rsrctype>conference_proceedings</rsrctype><prefilter>conference_proceedings</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Adaptation models</topic><topic>Artificial neural networks</topic><topic>Computational modeling</topic><topic>Computer Vision</topic><topic>Explainable AI</topic><topic>Fruit and Vegetable Classification</topic><topic>MLOps</topic><topic>Object detection</topic><topic>Pipelines</topic><topic>Production</topic><topic>Real-time systems</topic><topic>Robustness</topic><topic>Scalability</topic><topic>XAI</topic><toplevel>online_resources</toplevel><creatorcontrib>Iriarte, Francisco J.</creatorcontrib><creatorcontrib>Ortiz, Miguel E.</creatorcontrib><creatorcontrib>Unzueta, Luis</creatorcontrib><creatorcontrib>Martinez, Javier</creatorcontrib><creatorcontrib>Zaldivar, Javier</creatorcontrib><creatorcontrib>Arganda-Carreras, Ignacio</creatorcontrib><collection>IEEE Electronic Library (IEL) Conference Proceedings</collection><collection>IEEE Proceedings Order Plan All Online (POP All Online) 1998-present by volume</collection><collection>IEEE Xplore All Conference Proceedings</collection><collection>IEEE Xplore</collection><collection>IEEE Proceedings Order Plans (POP All) 1998-Present</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Iriarte, Francisco J.</au><au>Ortiz, Miguel E.</au><au>Unzueta, Luis</au><au>Martinez, Javier</au><au>Zaldivar, Javier</au><au>Arganda-Carreras, Ignacio</au><format>book</format><genre>proceeding</genre><ristype>CONF</ristype><atitle>Multi-Level XAI-Driven MLOps Pipeline for the Adjustment of Fruit and Vegetable Classifiers</atitle><btitle>2024 IEEE 12th International Conference on Intelligent Systems (IS)</btitle><stitle>IS</stitle><date>2024-08-29</date><risdate>2024</risdate><spage>1</spage><epage>6</epage><pages>1-6</pages><eissn>2767-9802</eissn><eisbn>9798350350982</eisbn><abstract>In this paper, we present a machine learning operations (MLOps) pipeline that exploits explainable artificial intelligence (XAI) to adjust deep neural network (DNN)-based fruit and vegetable image classifiers to data observed at super-market self-checkouts. DNNs are currently the most successful AI models for several automation tasks, including fruit and vegetable classification. However, adjusting them to on-site observations to avoid model drift is challenging, as they work as black boxes that take inputs and return results without showing the processes that lead to their responses. State-of-the-art XAI techniques could help mitigate this problem, but integrating them in MLOps pipelines is also challenging due to their high computational cost. We propose multi-task B-cos (MTBC) DNNs integrated at different pipeline levels to obtain automated information about the system's on-site performance. MTBC DNNs allow indentification of fruit and vegetable types with contextual attributes in real time and generation of contribution maps that highlight task-relevant features for enhanced model adjustment. Experimental results show that MTBC DNNs obtain similar accuracy and performance to the equivalent baseline DNNs for fruit and vegetable classification but with the added benefit of explainability, supporting MLOps processes such as drift detection and exploratory data analysis.</abstract><pub>IEEE</pub><doi>10.1109/IS61756.2024.10705202</doi></addata></record> |
fulltext | fulltext_linktorsrc |
identifier | EISSN: 2767-9802 |
ispartof | 2024 IEEE 12th International Conference on Intelligent Systems (IS), 2024, p.1-6 |
issn | 2767-9802 |
language | eng |
recordid | cdi_ieee_primary_10705202 |
source | IEEE Xplore All Conference Series |
subjects | Adaptation models Artificial neural networks Computational modeling Computer Vision Explainable AI Fruit and Vegetable Classification MLOps Object detection Pipelines Production Real-time systems Robustness Scalability XAI |
title | Multi-Level XAI-Driven MLOps Pipeline for the Adjustment of Fruit and Vegetable Classifiers |
url | http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-25T15%3A26%3A58IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-ieee_CHZPO&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=proceeding&rft.atitle=Multi-Level%20XAI-Driven%20MLOps%20Pipeline%20for%20the%20Adjustment%20of%20Fruit%20and%20Vegetable%20Classifiers&rft.btitle=2024%20IEEE%2012th%20International%20Conference%20on%20Intelligent%20Systems%20(IS)&rft.au=Iriarte,%20Francisco%20J.&rft.date=2024-08-29&rft.spage=1&rft.epage=6&rft.pages=1-6&rft.eissn=2767-9802&rft_id=info:doi/10.1109/IS61756.2024.10705202&rft.eisbn=9798350350982&rft_dat=%3Cieee_CHZPO%3E10705202%3C/ieee_CHZPO%3E%3Cgrp_id%3Ecdi_FETCH-ieee_primary_107052023%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_id=info:pmid/&rft_ieee_id=10705202&rfr_iscdi=true |