<div dir="ltr"><div dir="ltr" class="gmail_signature" data-smartmail="gmail_signature"><div dir="ltr"><div dir="ltr"><div id="gmail-:v4" class="gmail-Ar gmail-Au gmail-Ao"><div id="gmail-:824" class="gmail-Am gmail-Al editable gmail-LW-avf gmail-tS-tW gmail-tS-tY" aria-label="Message Body" role="textbox" aria-multiline="true" tabindex="1" style="direction:ltr;min-height:416px"><div dir="ltr" class="gmail_signature"><div dir="ltr"><div dir="ltr"><div id="gmail-:vs" class="gmail-Ar gmail-Au gmail-Ao"><div id="gmail-:7ae" class="gmail-Am gmail-Al editable gmail-LW-avf gmail-tS-tW gmail-tS-tY" aria-label="Message Body" role="textbox" aria-multiline="true" tabindex="1" style="direction:ltr;min-height:416px"><div dir="ltr" class="gmail_signature"><div dir="ltr"><div dir="ltr"><span id="gmail-docs-internal-guid-3b876413-7fff-eb06-c763-9cda732ea7e5"><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">We cordially invite those interested to our CVPR2022 virtual tutorial on </span><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-weight:700;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">Performance Measures in Visual Detection and Their Optimization </span><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">to be held online on 30 June 2022. </span></p><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">Tutorial Website: </span><a href="https://sites.google.com/view/performance-measures-cvpr2022/" style="text-decoration-line:none"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;text-decoration-line:underline;vertical-align:baseline;white-space:pre-wrap">https://sites.google.com/view/performance-measures-cvpr2022/</span></a><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap"> </span></p><br><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-weight:700;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">About the Tutorial</span></p><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">Many vision applications require identifying objects and object-related information in images. Such identification can be performed at different levels of detail, which are addressed by different visual detection tasks such as “object detection” for identifying labels of objects and boxes bounding them, “keypoint detection” for finding keypoints on objects, “instance segmentation” for identifying the classes of objects and localizing them with masks, and “panoptic segmentation” for both semantic segmentation of background classes and instance segmentation of objects. Accurately evaluating performances of these methods is crucial for developing better solutions.</span></p><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">Accordingly, in this tutorial, we aim to extensively delve into the evaluation of visual detectors. Within the scope of our tutorial, we will first cover the basics of evaluating visual detectors in order to allow someone not familiar with visual detection to grasp the basics. Then, we will introduce the Localisation Recall Precision (LRP) Error [1,2] and present thorough comparative both theoretical and comparative analyses with Average Precision (AP) and Panoptic Quality (PQ) [3] on various visual detection tasks. Finally, we will discuss bridging the gap between training and evaluation by directly optimizing AP and LRP, which involves a non-differentiable ranking step that is difficult to optimize using conventional gradient-based methods.</span></p><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-weight:700;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">Program (in CST)</span></p><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">Date: 30 June</span></p><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-weight:700;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">11.00am-11:50noon -- Part I: The Basics of Evaluating Visual Detectors [50 min]: </span></p><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">- Motivation and Introduction on Visual Detection Tasks</span></p><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">- Performance Measures in Visual Detection: Average Precision, Panoptic Quality, Localization Recall Precision (LRP) </span></p><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">- Recent Advances: Probabilistic Detection Quality, AP-fixed, AP-pool, Boundary IoU, Optimal Correction Cost </span></p><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-weight:700;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">11:50am-12:00noon -- Break</span></p><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-weight:700;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">12:00noon -- 12:50pm -- Part II: An Analysis of Performance Measures and Localisation-Recall-Precision Error [50 min]: </span></p><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">- An Analysis of Performance Measures: Important features for a performance measure, evaluating AP and PQ in terms of important features</span></p><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">- Localisation-Recall-Precision Error: Definition, Analysis, Optimal LRP Error, s-LRP Curves, Theoretical and Empirical Comparison of LRP Error with AP and PQ</span></p><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-weight:700;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">12:50pm-01:00pm -- Break</span></p><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-weight:700;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">01:00pm-01:50pm -- Part III: Optimization of Performance Measures [50 min]: </span></p><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">- Identity Update to Optimize Ranking-based Loss Functions</span></p><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">- Average Precision Loss for Classification</span></p><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">- Average Localisation-Recall-Precision Loss for Object Detection</span></p><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">- Rank & Sort Loss for Object Detection and Instance Segmentation</span></p><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-weight:700;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">01:50pm-02:00pm -- Q&A</span></p><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">Please check our webpage for up-to-date program: <a href="https://sites.google.com/view/performance-measures-cvpr2022/">https://sites.google.com/view/performance-measures-cvpr2022/</a></span></p><br><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-weight:700;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">Participation Details</span></p><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">Participation in our tutorial will be via the CVPR2022 platform and therefore will require registration to the conference.</span></p><br><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-weight:700;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">Organizing Committee</span></p><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">Emre Akbas, Sinan Kalkan, Kemal Oksuz</span></p><br><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-weight:700;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">References</span></p><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">[1] K. Oksuz, B. C. Cam, S. Kalkan*, E. Akbas*, "One Metric to Measure them All: Localisation Recall Precision (LRP) for Evaluating Visual Detection Tasks", IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), in press, 2022. [Paper] [Code]</span></p><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">[2] K. Oksuz, B. C. Cam, E. Akbas, S. Kalkan, "Localization Recall Precision (LRP): A New Performance Metric for Object Detection", European Conference on Computer Vision (ECCV), pp. 521-537, Springer, 2018. [Paper] [Code]</span></p><br><p dir="ltr" style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;color:rgb(0,0,0);background-color:transparent;font-variant-numeric:normal;font-variant-east-asian:normal;vertical-align:baseline;white-space:pre-wrap">[3] Kirillov, A., He, K., Girshick, R., Rother, C., & Dollár, P. "Panoptic segmentation". IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 9404-9413), 2019.</span></p></span><br class="gmail-Apple-interchange-newline"><div><br></div>-- <br><div dir="ltr" class="gmail_signature"><div dir="ltr"><div dir="ltr"><span style="font-size:12.8px">Sinan KALKAN, Assoc. Prof. Dr.</span></div><div dir="ltr"><br style="font-size:12.8px"><span style="font-size:12.8px">Dept. of Computer Engineering</span><br style="font-size:12.8px"><span style="font-size:12.8px">Middle East Technical University</span><br style="font-size:12.8px"><span style="font-size:12.8px">Ankara, TURKEY</span><br style="font-size:12.8px"><span style="font-size:12.8px">Web: </span><span style="font-size:12.8px"><a href="https://ceng.metu.edu.tr/~skalkan/" target="_blank">https://ceng.metu.edu.tr/~skalkan/</a></span></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div>