<html><head><meta http-equiv="content-type" content="text/html; charset=utf-8"></head><body style="overflow-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;"><meta http-equiv="content-type" content="text/html; charset=utf-8"><div style="overflow-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;"><meta http-equiv="content-type" content="text/html; charset=utf-8"><div style="overflow-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;"><meta http-equiv="content-type" content="text/html; charset=utf-8"><div style="overflow-wrap: break-word; -webkit-nbsp-mode: space; line-break: after-white-space;"><div dir="ltr">Reiteramos la invitación.</div><div dir="ltr"><br></div><div dir="ltr">Saludos,</div><div dir="ltr"><br></div><div dir="ltr">Pablo</div><div dir="ltr"><br></div><div dir="ltr"><br></div><div dir="ltr">——————————</div><div dir="ltr"><br></div><div dir="ltr"><br></div><div dir="ltr">Estimados,</div><div dir="ltr"><br></div><div dir="ltr">Es un placer invitarlos a la defensa de la tesis de doctorado en ingeniería eléctrica de Matías Tailanián, que tuve el placer de dirigir junto a Alvaro Pardo. La defensa tendrá lugar en Facultad de Ingeniería, y será además transmitida por zoom. </div><div dir="ltr"><br></div><div dir="ltr">Van todos los detalles abajo.</div><div dir="ltr"><br></div><div dir="ltr"><div dir="ltr">Saludos,</div><div dir="ltr"><br></div><div dir="ltr">Pablo</div><div dir="ltr"><br></div><div dir="ltr"></div></div><div dir="ltr">—————————</div><div dir="ltr"><br></div><div dir="ltr">Fecha: martes 17 de setiembre, 8h@Uruguay</div><div dir="ltr"><br></div><div dir="ltr">Lugar:</div><div dir="ltr"><br></div><div dir="ltr">- Facultad de Ingeniería, salón 705 (piso 7, cuerpo central). Habrá un brindis a continuación en Instituto de Ingeniería Eléctrica.</div><div dir="ltr"><div class="gmail_quote">
<br>- Zoom : <br><br><a href="https://salavirtual-udelar.zoom.us/j/2165307614?pwd=c1EvSnlUUFg0TDlKUDVRd3lKOG01Zz09">https://salavirtual-udelar.zoom.us/j/2165307614?pwd=c1EvSnlUUFg0TDlKUDVRd3lKOG01Zz09</a><br>Meeting ID: 216 530 7614<br>Passcode: pwd-AB2021</div><div class="gmail_quote"><br><br></div><div class="gmail_quote"><br></div><div class="gmail_quote">Título: <b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">Deep Image Generative Modeling and Statistical Testing for Industrial Anomaly Detection</b><br>
<br>
Abstract:</div><div class="gmail_quote"><br><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">This thesis addresses the challenge of </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">anomaly detection</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"> in images, for </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">industrial applications.</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"> It explores advanced methodologies employing both </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">classical image processing</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"> techniques and modern </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">generative modeling</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"> approaches, specifically focusing on </span><i data-stringify-type="italic" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">Normalizing Flows</i><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"> and </span><i data-stringify-type="italic" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">Diffusion Models</i><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">.</span><span aria-label="" class="c-mrkdwn__br" data-stringify-type="paragraph-break" style="box-sizing: inherit; height: 8px; display: block; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"></span><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">As anomalies are rare by definition, collecting normal samples is generally easier and more feasible in industrial settings than acquiring comprehensive datasets with labeled anomalies. Therefore, the focus of this research is on </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">unsupervised methods</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">, and </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">one-class methods</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">, where the idea is to model the "normality" and detect anomalies as everything that deviates from this model.</span><span aria-label="" class="c-mrkdwn__br" data-stringify-type="paragraph-break" style="box-sizing: inherit; height: 8px; display: block; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"></span><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">Initially, a multi-scale anomaly detection method based on classical image processing techniques is proposed, leveraging an </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"><i data-stringify-type="italic" style="box-sizing: inherit;">a contrario</i></b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"> approach to control the number of false alarms. Subsequently, a novel method called </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">U-Flow</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"> is introduced, which employs a U-shaped architecture in </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">Normalizing Flows</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"> to achieve anomaly detection with automatic thresholding. Then, this thesis further explores the use of </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">Diffusion Models</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"> for anomaly detection, presenting the </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">Diffusion Anomaly Detection (DAD)</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"> method. This work incorporates </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">score-based generative models</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"> and </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">inpainting</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"> techniques to refine anomaly detection capabilities. Additionally, a new diffusion-based method called </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">RIFA (Random Inpainting For Anomaly detection)</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"> is proposed as a completely unsupervised alternative. Finally, the techniques and knowledge gained from Diffusion Models are applied to a completely different application: </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">counter-forensics</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">.</span><span aria-label="" class="c-mrkdwn__br" data-stringify-type="paragraph-break" style="box-sizing: inherit; height: 8px; display: block; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"></span><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">Throughout the whole thesis, a special emphasis is placed on bridging the gap between theoretical research and practical industrial applications, setting the theoretical foundations for obtaining automatic segmentations of anomalies, by performing </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">statistical tests</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"> and controlling the number of false alarms using the </span><i data-stringify-type="italic" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">a contrario</i><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"> framework.</span><span aria-label="" class="c-mrkdwn__br" data-stringify-type="paragraph-break" style="box-sizing: inherit; height: 8px; display: block; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"></span><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">Experimental results on standard datasets validate the effectiveness of the proposed methods, highlighting substantial performance gains in some cases. The final chapter applies the best-performing method to </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">two industrial problems: quality control in manufacturing leather samples</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"> for the upholstery industry, and </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">defect detection in fruits</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">, demonstrating its practical viability and impact on improving quality control processes in these industries.</span><span aria-label="" class="c-mrkdwn__br" data-stringify-type="paragraph-break" style="box-sizing: inherit; height: 8px; display: block; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"></span><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">In addition, this research contributes to the open-source community with several </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">code repositories</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;"> and has resulted in </span><b data-stringify-type="bold" style="box-sizing: inherit; font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">four published papers so far</b><span style="font-family: Slack-Lato, Slack-Fractions, appleLogo, sans-serif; font-variant-ligatures: common-ligatures; orphans: 2; widows: 2; text-decoration-thickness: initial; text-decoration-style: initial; text-decoration-color: initial;">, and hopefully, more will follow. Future work will particularly focus on improving likelihood estimation with Diffusion Models and expanding its applicability to other industrial domains.</span></div><div class="gmail_quote"><div style="orphans: 2; widows: 2;"><font face="Slack-Lato, Slack-Fractions, appleLogo, sans-serif"><br></font></div><div style="orphans: 2; widows: 2;"><font face="Slack-Lato, Slack-Fractions, appleLogo, sans-serif"><br></font></div><div style="orphans: 2; widows: 2;"><font face="Slack-Lato, Slack-Fractions, appleLogo, sans-serif"><br></font></div>
------------------------------------<br>
<br>
Tribunal:</div><div class="gmail_quote"><br></div><div class="gmail_quote">Matías Di Martino (Universidad Católica del Uruguay)<br>Thibaud Ehret (Ecole Normale Supérieure Paris-Saclay)</div><div class="gmail_quote">Jean-Michel Morel (City University of Hong Kong)</div><div class="gmail_quote">Gregory Randall (Universidad de la República)</div><div class="gmail_quote"><br></div><div class="gmail_quote"><br>
Directores de tesis:</div><div class="gmail_quote"><br>Pablo Musé (Universidad de la República)</div><div class="gmail_quote">Alvaro Pardo (Universidad Católica del Uruguay)</div><div class="gmail_quote">
<br>
————————————————————</div><div class="gmail_quote"><br></div></div><br><div><br></div></div></div><br><div>
----------------------------------------------<br>Pablo Musé, PhD<br>Full Professor of Signal Processing<br>Universidad de la República <br>Uruguay <br>+598 27110974<br>iie.fing.edu.uy/~pmuse/
</div>
<br></div></body></html>