BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Department of Electrical and Computer Engineering (HKUECE) 電機與計算機工程系 - ECPv6.16.0//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:Department of Electrical and Computer Engineering (HKUECE) 電機與計算機工程系
X-ORIGINAL-URL:https://ece.hku.hk
X-WR-CALDESC:Events for Department of Electrical and Computer Engineering (HKUECE) 電機與計算機工程系
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-Robots-Tag:noindex
X-PUBLISHED-TTL:PT1H
BEGIN:VTIMEZONE
TZID:Asia/Hong_Kong
BEGIN:STANDARD
TZOFFSETFROM:+0800
TZOFFSETTO:+0800
TZNAME:HKT
DTSTART:20220101T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=Asia/Hong_Kong:20231222T100000
DTEND;TZID=Asia/Hong_Kong:20231222T110000
DTSTAMP:20260513T073709
CREATED:20231213T064055Z
LAST-MODIFIED:20250114T075508Z
UID:17878-1703239200-1703242800@ece.hku.hk
SUMMARY:RPG Seminar – CL-NeRF: Continual Learning of Neural Radiance Fields for Evolving Scene Representation
DESCRIPTION:Existing methods for adapting Neural Radiance Fields (NeRFs) to scene changes require extensive data capture and model retraining\, which is both time-consuming and labor-intensive. In this paper\, we tackle the challenge of efficiently adapting NeRFs to real-world scene changes over time using a few new images while retaining the memory of unaltered areas\, focusing on the continual learning aspect of NeRFs. To this end\, we propose CL-NeRF\, which consists of two key components: a lightweight expert adaptor for adapting to new changes and evolving scene representations and a conflict-aware knowledge distillation learning objective for memorizing unchanged parts. We also present a new benchmark for evaluating Continual Learning of NeRFs with comprehensive metrics. Our extensive experiments demonstrate that CL-NeRF can synthesize high-quality novel views of both changed and unchanged regions with high training efficiency\, surpassing existing methods in terms of reducing forgetting and adapting to changes. \nZoom Link :\nhttps://hku.zoom.us/j/3164938755 \nBiography of the speaker:\n\nXiuzhe Wu received her B.Eng. and M.Eng degrees at Tongji University. Currently\, she is pursuing Ph.D. degree at the University of Hong Kong\, advised by Dr. Xiaojuan Qi. Her research interests focus on 3D vision. \nAll are welcome.
URL:https://ece.hku.hk/events/rpg-seminar-cl-nerf-continual-learning-of-neural-radiance-fields-for-evolving-scene-representation/
CATEGORIES:Seminar
ATTACH;FMTTYPE=image/jpeg:https://ece.hku.hk/wp-content/uploads/2024/11/rpg-seminar.jpg
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Asia/Hong_Kong:20231222T100000
DTEND;TZID=Asia/Hong_Kong:20231222T110000
DTSTAMP:20260513T073709
CREATED:20231218T080324Z
LAST-MODIFIED:20250114T074936Z
UID:17888-1703239200-1703242800@ece.hku.hk
SUMMARY:RPG Seminar – Can 3D Vision-Language Models Truly Understand Natural Language?
DESCRIPTION:Rapid advancements in 3D vision-language (3D-VL) tasks\, such as 3D Visual Question Answering (3D-VQA) and 3D Visual Grounding (3D-VG)\, have opened up new avenues for human interaction with embodied agents or robots using natural language. Despite this progress\, we find a notable limitation: existing 3D-VL models exhibit heightened sensitivity to language input\, struggling to handle sentences with minor stylistic changes. This observation raises the critical question: “Can 3D vision-language models truly understand natural language?” We first propose a language robustness task for systematically assessing 3D-VL models across various tasks\, benchmarking their performance when presented with different language style variants. We propose the first 3D Language Robustness Dataset\, designed based on the characteristics of human language\, to facilitate the systematic study of robustness. Our comprehensive evaluation uncovers a significant drop in the performance of all existing models across various 3D-VL tasks. Even the state-of-the-art 3D-LLM fails on some variances. Further in-depth analysis suggests that the existing model fails to align the feature space\, which also stems from the low diversity of the existing dataset. We also propose a plug-and-play training free pre-alignment module driven by LLM\, which can improve language robustness. The data will be available to facilitate further research. \nZoom Link :\nhttps://hku.zoom.us/j/96527514816\nMeeting ID: 965 2751 4816 \nBiography of the speaker:\n\nWeipeng DENG received his B.Eng. of Computer Science from South China University of Technology. Currently\, he is pursuing his Ph.D. degree at the University of Hong Kong. His work is focused on how to leverage language to assist vision-language system. His research interests include deep learning\, large language model\, prompt\, vision-language model. \nAll are welcome.
URL:https://ece.hku.hk/events/rpg-seminar-can-3d-vision-language-models-truly-understand-natural-language/
CATEGORIES:Seminar
ATTACH;FMTTYPE=image/jpeg:https://ece.hku.hk/wp-content/uploads/2024/11/rpg-seminar.jpg
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Asia/Hong_Kong:20231222T110000
DTEND;TZID=Asia/Hong_Kong:20231222T120000
DTSTAMP:20260513T073709
CREATED:20231213T063326Z
LAST-MODIFIED:20250114T075645Z
UID:17876-1703242800-1703246400@ece.hku.hk
SUMMARY:RPG Seminar – Video Demoireing with Relation-Based Temporal Consistency
DESCRIPTION:Moire patterns\, appearing as color distortions\, severely degrade the image and video qualities when filming a screen with digital cameras. Considering the increasing demands for capturing videos\, we study how to remove such undesirable moire patterns in videos\, namely video demoireing. To this end\, we introduce the first hand-held video demoireing dataset with a dedicated data collection pipeline to ensure spatial and temporal alignments of captured data. Further\, a baseline video demoireing model with implicit feature space alignment and selective feature aggregation is developed to leverage complementary information from nearby frames to improve frame-level video demoireing. More importantly\, we propose a relation-based temporal consistency loss to encourage the model to learn temporal consistency priors directly from ground-truth reference videos\, which facilitates producing temporally consistent predictions and effectively maintains frame-level qualities. Extensive experiments manifest the superiority of our model. \nZoom Link :\nhttps://hku.zoom.us/j/95468941119?pwd=UVpSUGowM3F2OXlwUzN6ajhvWHVQUT09\nMeeting ID: 954 6894 1119\nPassword: 336484 \nBiography of the speaker:\n\nPeng Dai is a fourth-year Ph.D. candidate at CVMI Lab\, Department of Electrical and Electronic Engineering\, The University of Hong Kong (HKU). His supervisor is Dr. Xiaojuan Qi. His research interest includes computer vision and computer graphics. \nAll are welcome.
URL:https://ece.hku.hk/events/rpg-seminar-video-demoireing-with-relation-based-temporal-consistency/
CATEGORIES:Seminar
ATTACH;FMTTYPE=image/jpeg:https://ece.hku.hk/wp-content/uploads/2024/11/rpg-seminar.jpg
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Asia/Hong_Kong:20231222T110000
DTEND;TZID=Asia/Hong_Kong:20231222T120000
DTSTAMP:20260513T073709
CREATED:20231214T090837Z
LAST-MODIFIED:20250114T075053Z
UID:17884-1703242800-1703246400@ece.hku.hk
SUMMARY:RPG Seminar – Single-pulse optogenetic perturbation of thalamo-cortical networks reveals rsfMRI architecture
DESCRIPTION:Resting-state fMRI (rsfMRI) has emerged as the most valuable\, non-invasive imaging technique to map long-range\, brain-wide functional connectivity networks. Recently\, dynamic rsfMRI network segregation and integration have been shown to facilitate and modulate diverse cognitive functions. Such dynamics are structural-functional hierarchically organized and have been observed across a range of temporal scales\, spanning from days or hours\, to minutes. Converging studies postulate that critical transient states and their transition processes subserve such functional architecture of rsfMRI networks. \nHowever\, previous studies using either rsfMRI measurements only (i.e.\, during awake/sleep state) or task-based fMRI with conventional stimulation paradigms (i.e.\, block-designed/pulse trains) fail to clearly dissect transient/seconds-level reorganization of rsfMRI architecture upon stimulations. Here\, we propose to implement a single-pulse stimulation design to exert minimum influence on spontaneous activities (e.g.\, occurrence/intensity of spontaneous neural events) while modulating rsfMRI transient states and transition processes. \nZoom Link :\nhttps://hku.zoom.us/j/7179232708?omn=92069441167 \nBiography of the speaker:\n\nLinshan Xie obtained her B.Eng degree in Electronic Science and Technology from the University of Electronic Science And Technology Of China in 2020. She is now pursuing PhD in the Department of Electrical and Electronic Engineering at the University of Hong Kong under the supervision of Prof. Ed X Wu and Dr. Alex T.L. Leong. Her research interests focus on fMRI application in neuroscience\, including task-based (optogenetic) and resting-state fMRI acquisition and analysis. \nAll are welcome.
URL:https://ece.hku.hk/events/rpg-seminar-single-pulse-optogenetic-perturbation-of-thalamo-cortical-networks-reveals-rsfmri-architecture/
CATEGORIES:Seminar
ATTACH;FMTTYPE=image/jpeg:https://ece.hku.hk/wp-content/uploads/2024/11/rpg-seminar.jpg
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Asia/Hong_Kong:20231222T140000
DTEND;TZID=Asia/Hong_Kong:20231222T150000
DTSTAMP:20260513T073709
CREATED:20231218T080705Z
LAST-MODIFIED:20250114T074904Z
UID:17889-1703253600-1703257200@ece.hku.hk
SUMMARY:RPG Seminar – Neural computing in random resistive memory
DESCRIPTION:Arrays\, graphs\, and point sets are fundamental data organization forms in mathematics\, with significant applications in scientific computing and engineering problems like signal processing\, molecular discovery\, and 3D vision. However\, the progress of digital circuitry\, limited by Moore’s Law\, has slowed down\, impeding the advancement of these problems. The increasing data and computational demands have amplified the von Neumann bottleneck in traditional computing systems\, necessitating urgent improvements in energy efficiency\, latency\, and computational capacity. Novel memory-based computing paradigms\, such as resistive memories\, have made progress in addressing these challenges. However\, resistive memories suffer from high power consumption and delay during write operations\, due to the need for high voltages to form conductive filaments. Randomness in filament shape and size also introduces inaccuracies. To overcome these issues\, we propose a co-design approach that minimizes programming/write operations in machine learning systems handling arrays\, graphs\, and point sets. This method improves system efficiency and guarantees performance. Validation on tasks involving arrays\, graphs\, and point sets demonstrates superior task performance and significant energy reduction compared to traditional computing systems. These research opens doors to future in-memory computing acceleration system design. \nZoom Link :\nhttps://hku.zoom.us/j/98133426720 \nBiography of the speaker:\n\nShaocong Wang is a fourth-year Ph.D. candidate at IMSC lab\, Department of Electrical and Electronic Engineering\, The University of Hong Kong (HKU). His supervisor is  Dr. Zhongrui Wang. His research interest includes computing-in-memory and analogue computing. \nAll are welcome.
URL:https://ece.hku.hk/events/rpg-seminar-neural-computing-in-random-resistive-memory/
CATEGORIES:Seminar
ATTACH;FMTTYPE=image/jpeg:https://ece.hku.hk/wp-content/uploads/2024/11/rpg-seminar.jpg
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Asia/Hong_Kong:20231222T150000
DTEND;TZID=Asia/Hong_Kong:20231222T160000
DTSTAMP:20260513T073709
CREATED:20231218T075938Z
LAST-MODIFIED:20250114T075012Z
UID:17887-1703257200-1703260800@ece.hku.hk
SUMMARY:RPG Seminar – Large-scale single-cell morphological profiling on a spinning arrayed optofluidic platform
DESCRIPTION:Morphological profiling in imaging microscopy can reveal biological cells characteristics. However\, current image-based assay relays heavily on costly and extensive amount of fluorescent antibodies and is limited by the trade-off between the image resolution and scalability of the experiment. Therefore\, here we presented a spinning arrayed optofluidic platform to achieve live cell\, label free\, large-scale and high-resolution imaging. We demonstrated the stability of the system can achieve long-term imaging monitoring for at least 30 mins and further validated that the platform shows minimal effects on cells health based of a panel of cell-health assays. The application of the platform and morphological profiling was demonstrated through a drug screening analysis on two ling cancer cell lines with four drugs having five different concentrations\, which we observed that drug-concentration applied to lung-cancer cells showed obvious morphological profile shifts. Our profiling technique also showed the potential on understanding the viral entry mechanisms and impact on label-free cell morphology from quantitative phase imaging on CRISPR-based cell-cell fusion assay involving human ACE2 receptor cells and SARS-CoV-2 spike protein expression cells. The results show new and valuable insights on drug treatment and disease- and gene-related phenotypes. \nZoom Link :\nhttps://hku.zoom.us/j/95494140198?pwd=WEtVcEtFRjdzdGdacmFReFZLWkhmUT09\nMeeting ID: 954 9414 0198\nPassword: 217678 \nBiography of the speaker:\n\nVictor Wong received the BEng in medical engineering in 2019 and is currently pursuing the Ph.D. degree with the Department of Electrical and Electronic Engineering\, The University of Hong Kong (HKU). His research interests focus on ultrafast imaging\, single cell analysis\, and morphological profiling. He was also awarded the Research Postgraduate Student Innovation Award in 2022/23 established by Graduate School and the Technology Transfer Office. \nAll are welcome.
URL:https://ece.hku.hk/events/rpg-seminar-large-scale-single-cell-morphological-profiling-on-a-spinning-arrayed-optofluidic-platform/
CATEGORIES:Seminar
ATTACH;FMTTYPE=image/jpeg:https://ece.hku.hk/wp-content/uploads/2024/11/rpg-seminar.jpg
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Asia/Hong_Kong:20231222T160000
DTEND;TZID=Asia/Hong_Kong:20231222T170000
DTSTAMP:20260513T073709
CREATED:20231218T081037Z
LAST-MODIFIED:20250114T074745Z
UID:17890-1703260800-1703264400@ece.hku.hk
SUMMARY:RPG Seminar – High Speed Quantitative Phase and Polarization Microscopy for Zebrafish In Vivo Blood Cell Imaging
DESCRIPTION:Quantitative phase and polarization microscope (QPPI) is a useful label-free microscopic tool in revealing the quantitative biophysical information of cells\, such as cell shape\, dry mass and their subcellular distribution. However\, the current existing QPPI system could only offer a low imaging speed which is not sufficient to observe the dynamic activities of the live animal samples\, which is very important in understanding the disease developments and drug treatment effects. One example would be on the zebrafish leukaemia model. Fast imaging speed with high content images could be able to help monitor the in vivo blood flowing cells\, which provide the dynamic information of the cancer situations and drug effects after the different treatments. To boost the imaging speed of the QPPI system\, one possible solution is to combine the QPPI imaging technique with the existing developed ultrafast imaging system which is the Free-space Angular-chirp-enhanced Delay (FACED). This could help boost the imaging speed up to kHz which is sufficient to observe any zebrafish blood cell movements and reveal more information on the leukaemia as well as the effect of drugs. \nZoom Link :\nhttps://hku.zoom.us/j/5701095284?pwd=SUJYUFg1Q2pybUZhNE5WbnBqUXBSdz09 \nBiography of the speaker:\n\nRicky Hui received the BEng in Biomedical Engineering (BME) in 2022 in The University of Hong Kong (HKU) and is currently pursuing the MPhil Degree with the Department of Electrical and Electronic Engineering\, The University of Hong Kong (HKU). His research interests focus on ultrafast imaging\, volumetric imaging\, and quantitative label-free imaging techniques. \nAll are welcome.
URL:https://ece.hku.hk/events/rpg-seminar-high-speed-quantitative-phase-and-polarization-microscopy-for-zebrafish-in-vivo-blood-cell-imaging/
CATEGORIES:Seminar
ATTACH;FMTTYPE=image/jpeg:https://ece.hku.hk/wp-content/uploads/2024/11/rpg-seminar.jpg
END:VEVENT
END:VCALENDAR