Publications without Fulltext
Permanent URI for this collectionhttps://hdl.handle.net/20.500.14288/3
Browse
1737 results
Filters
Advanced Search
Filter by
Settings
Search Results
Publication Metadata only Self-supervised object-centric learning for videos(Neural information processing systems foundation, 2023) Xie, Weidi; Department of Computer Engineering; Aydemir, Görkay; Güney, Fatma; Department of Computer Engineering; Graduate School of Sciences and Engineering; College of EngineeringUnsupervised multi-object segmentation has shown impressive results on images by utilizing powerful semantics learned from self-supervised pretraining. An additional modality such as depth or motion is often used to facilitate the segmentation in video sequences. However, the performance improvements observed in synthetic sequences, which rely on the robustness of an additional cue, do not translate to more challenging real-world scenarios. In this paper, we propose the first fully unsupervised method for segmenting multiple objects in real-world sequences. Our object-centric learning framework spatially binds objects to slots on each frame and then relates these slots across frames. From these temporally-aware slots, the training objective is to reconstruct the middle frame in a high-level semantic feature space. We propose a masking strategy by dropping a significant portion of tokens in the feature space for efficiency and regularization. Additionally, we address over-clustering by merging slots based on similarity. Our method can successfully segment multiple instances of complex and high-variety classes in YouTube videos.Publication Metadata only Validating digital traces with survey data: the use case of religiosity(Association for Computing Machinery, Inc, 2024) Department of Sociology;Department of Mathematics; Yörük, Erdem; Atsızelti, Şükrü; Yardı, Melih Can; Duruşan, Fırat; Etgü, Tolga; College of Social Sciences and Humanities; Graduate School of Social Sciences and Humanities; College of SciencesThis paper tests the validity of a digital trace database (Politus) obtained from Twitter, with a recently conducted representative social survey, focusing on the use case of religiosity in Turkey. Religiosity scores in the research are extracted using supervised machine learning under the Politus project. The validation analysis depends on two steps. First, we compare the performances of two alternative tweet-To-user transformation strategies, and second, test for the impact of resampling via the MRP technique. Estimates of the Politus are examined at both aggregate and region-level. The results are intriguing for future research on measuring public opinion via social media data.Publication Metadata only Vilma: a zero-shot benchmark for linguıstic and temporal grounding in video-language models(International Conference on Learning Representations, ICLR, 2024) Pedrotti, Andrea; Dogan, Mustafa; Cafagna, Michele; Parcalabescu, Letitia; Calixto, Iacer; Frank, Anetteh; Gatt, Albert; Department of Electrical and Electronics Engineering; Kesen, İlker; Erdem, Aykut; Department of Electrical and Electronics Engineering; Koç Üniversitesi İş Bankası Yapay Zeka Uygulama ve Araştırma Merkezi (KUIS AI)/ Koç University İş Bank Artificial Intelligence Center (KUIS AI); Graduate School of Sciences and Engineering; College of EngineeringWith the ever-increasing popularity of pretrained Video-Language Models (VidLMs), there is a pressing need to develop robust evaluation methodologies that delve deeper into their visio-linguistic capabilities. To address this challenge, we present VILMA), a task-agnostic benchmark that places the assessment of fine-grained capabilities of these models on a firm footing. Task-based evaluations, while valuable, fail to capture the complexities and specific temporal aspects of moving images that VidLMs need to process. Through carefully curated counterfactuals, VILMA offers a controlled evaluation suite that sheds light on the true potential of these models, as well as their performance gaps compared to human-level understanding. VILMA also includes proficiency tests, which assess basic capabilities deemed essential to solving the main counterfactual tests. We show that current VidLMs' grounding abilities are no better than those of vision-language models which use static images. This is especially striking once the performance on proficiency tests is factored in. Our benchmark serves as a catalyst for future research on VidLMs, helping to highlight areas that still need to be explored.Publication Metadata only Contrast improvement through a Generative Adversarial Network (GAN) by utilizing a dataset obtained from a line-scanning confocal microscope(SPIE, 2024) Department of Physics; Kiraz, Alper; Morova, Berna; Bavili, Nima; Ketabchi, Amir Mohammad; Department of Physics; Koç University Research Center for Translational Medicine (KUTTAM) / Koç Üniversitesi Translasyonel Tıp Araştırma Merkezi (KUTTAM); College of Sciences; Graduate School of Sciences and EngineeringConfocal microscopy offers enhanced image contrast and signal-to-noise ratio compared to wide-field illumination microscopy, achieved by effectively eliminating out-of-focus background noise. In our study, we initially showcase the functionality of a line-scanning confocal microscope aligned through the utilization of a Digital Light Projector (DLP) and a rolling shutter CMOS camera. In this technique, a sequence of illumination lines is projected onto a sample using a DLP and focusing objective (50X, NA=0.55). The reflected light is imaged with the camera. Line-scanning confocal imaging is accomplished by synchronizing the illumination lines with the rolling shutter of the sensor, leading to a substantial enhancement of approximately 50% in image contrast. Subsequently, this setup is employed to create a dataset comprising 500 pairs of images of paper tissue. This dataset is employed for training a Generative Adversarial Network (cGAN). Roughly 45% contrast improvement was measured in the test images for the trained network, in comparison to the ground-truth images.Publication Metadata only Multi-scale deformable alignment and content-adaptive inference for flexible-rate bi-directional video compression(IEEE Computer Society, 2023) Department of Electrical and Electronics Engineering; Yılmaz, Mustafa Akın; Ulaş, Ökkeş Uğur; Tekalp, Ahmet Murat; Department of Electrical and Electronics Engineering; Graduate School of Sciences and Engineering; College of EngineeringThe lack of ability to adapt the motion compensation model to video content is an important limitation of current end-to-end learned video compression models. This paper advances the state-of-the-art by proposing an adaptive motion-compensation model for end-to-end rate-distortion optimized hierarchical bi-directional video compression. In particular, we propose two novelties: i) a multi-scale deformable alignment scheme at the feature level combined with multi-scale conditional coding, ii) motion-content adaptive inference. In addition, we employ a gain unit, which enables a single model to operate at multiple rate-distortion operating points. We also exploit the gain unit to control bit allocation among intra-coded vs. bi-directionally coded frames by fine tuning corresponding models for truly flexible-rate learned video coding. Experimental results demonstrate state-of-the-art rate-distortion performance exceeding those of all prior art in learned video coding1.Publication Metadata only Virtual collaboration tools for mixed-ability workspaces: a cross disability solidarity case from Turkey(Assoc Computing Machinery, 2023) Department of Media and Visual Arts; Yıldız, Zeynep; Subaşı, Özge; Department of Media and Visual Arts; Graduate School of Social Sciences and HumanitiesA growing body of literature on mixed-ability teams within HCI investigates how disabled and non-disabled people collaborate. Still, how diferent disabilities can interact in a mixed-ability team is underexplored, especially for long commitments and in non-western contexts. As an emerging perspective in accessibility studies in HCI, disability justice emphasizes the importance of cross-disability collaborations. Collaborative access, interdependence, and crossdisability dialogue are keys to building accessible mixed-ability interactions. We conducted ten in-depth interviews with the members of a unique mixed-ability team (which includes people with diferent physical disabilities) using the same workspace with crossdisability interactions in Turkey. We aim to understand the requirements for an accessible mixed-ability virtual workspace and to identify practical design considerations for cross-disability solidarityoriented virtual collaboration tools. To ensure equal access in virtual workspaces, we suggest implications for centering collective access, balancing external power dynamics, and supporting language and cultural diversities.Publication Metadata only Ris-aided angular-based hybrid beamforming design in mmwave massive mimo systems(IEEE, 2022) Koc, Asil; Tho Le-Ngoc; Department of Electrical and Electronics Engineering; Yıldırım, İbrahim; Başar, Ertuğrul; Department of Electrical and Electronics Engineering; Graduate School of Sciences and Engineering; College of EngineeringThis paper proposes a reconfigurable intelligent surface (RIS)-aided and angular-based hybrid beamforming (AB-HBF) technique for the millimeter wave (mmWave) massive multiple-input multiple-output (MIMO) systems. The proposed RIS-AB-HBF architecture consists of three stages: (i) RF beam-former, (ii) baseband (BB) precoder/combiner, and (iii) RIS phase shift design. First, in order to reduce the number of RF chains and the channel estimation overhead, RF beamformers are designed based on the 3D geometry-based mmWave channel model using slow time-varying angular parameters of the channel. Second, a BB precoder/combiner is designed by exploiting the reduced-size effective channel seen from the BB stages. Then, the phase shifts of the RIS are adjusted to maximize the achievable rate of the system via the nature-inspired particle swarm optimization (PSO) algorithm. Illustrative simulation results demonstrate that the use of RISs in the AB-HBF systems has the potential to provide more promising advantages in terms of reliability and flexibility in system design.Publication Metadata only Prognostic value of isolated tumor cells in sentinel lymph nodes in low risk endometrial cancer: results from an international multi-institutional study(BMJ Publishing Group, 2023) Cucinella, Giuseppe; Schivardi, Gabriella; Zhou, Xun Clare; Alhilli, Mariam; Wallace, Sumer; Wohlmuth, Christoph; Baiocchi, Glauco; Tokgozoglu, Nedim; Raspagliesi, Francesco; Buda, Alessandro; Zanagnolo, Vanna; Zapardiel, Ignacio; Jagasia, Nisha; Giuntoli, Robert; Glickman, Ariel; Peiretti, Michele; Lanner, Maximilian; Chacon, Enrique; Di Guilmi, Julian; Pereira, Augusto; Laas-Faron, Enora; Fishman, Ami; Nitschmann, Caroline C.; Kurnit, Katherine; Moriarty, Kristen; Joehlin-Price, Amy; Lees, Brittany; Covens, Allan; De Brot, Louise; Bogani, Giorgio; Landoni, Fabio; Grassi, Tommaso; Paniga, Cristiana; Multinu, Francesco; De Vitis, Luigi Antonio; Hernández, Alicia; Mastroyannis, Spyridon; Ghoniem, Khaled; Chiantera, Vito; Shahi, Maryam; Fought, Angela J.; McGree, Michaela; Mariani, Andrea; Glaser, Gretchen; Taşkıran, Çağatay; School of MedicineObjective: The prognostic significance of isolated tumor cells (≤0.2 mm) in sentinel lymph nodes (SLNs) of endometrial cancer patients is still unclear. Our aim was to assess the prognostic value of isolated tumor cells in patients with low risk endometrial cancer who underwent SLN biopsy and did not receive adjuvant therapy. Outcomes were compared with node negative patients. Methods: Patients with SLNs-isolated tumor cells between 2013 and 2019 were identified from 15 centers worldwide, while SLN negative patients were identified from Mayo Clinic, Rochester, between 2013 and 2018. Only low risk patients (stage IA, endometrioid histology, grade 1 or 2) who did not receive any adjuvant therapy were included. Primary outcomes were recurrence free, non-vaginal recurrence free, and overall survival, evaluated with Kaplan-Meier methods. Results: 494 patients (42 isolated tumor cells and 452 node negative) were included. There were 21 (4.3%) recurrences (5 SLNs-isolated tumor cells, 16 node negative); recurrence was vaginal in six patients (1 isolated tumor cells, 5 node negative), and non-vaginal in 15 (4 isolated tumor cells, 11 node negative). Median follow-up among those without recurrence was 2.3 years (interquartile range (IQR) 1.1-3.0) and 2.6 years (IQR 0.6-4.2) in the SLN-isolated tumor cell and node negative patients, respectively. The presence of SLNs-isolated tumor cells, lymphovascular space invasion, and International Federation of Obstetrics and Gynecology (FIGO) grade 2 were significant risk factors for recurrence on univariate analysis. SLN-isolated tumor cell patients had worse recurrence free survival (p<0.01) and non-vaginal recurrence free survival (p<0.01) compared with node negative patients. Similar results were observed in the subgroup of patients without lymphovascular space invasion (n=480). There was no difference in overall survival between the two cohorts in the full sample and the subset excluding patients with lymphovascular space invasion. Conclusions: Patients with SLNs-isolated tumor cells and low risk profile, without adjuvant therapy, had a significantly worse recurrence free survival compared with node negative patients with similar risk factors, after adjusting for grade and excluding patients with lymphovascular space invasion. However, the presence of SLNs-isolated tumor cells was not associated with worse overall survival.Publication Metadata only A kernel-based multilayer perceptron framework to identify pathways related to cancer stages(Springer International Publishing Ag, 2023) Mokhtaridoost, Milad; Department of Industrial Engineering; Soleimanpoor, Marzieh; Gönen, Mehmet; Department of Industrial Engineering; Graduate School of Sciences and Engineering; College of EngineeringStandard machine learning algorithms have limited knowledge extraction capability in discriminating cancer stages based on genomic characterizations, due to the strongly correlated nature of high-dimensional genomic data. Moreover, activation of pathways plays a crucial role in the growth and progression of cancer from early-stage to latestage. That is why we implemented a kernel-based neural network framework that integrates pathways and gene expression data using multiple kernels and discriminates early- and late-stages of cancers. Our goal is to identify the relevant molecular mechanisms of the biological processes which might be driving cancer progression. As the input of developed multilayer perceptron (MLP), we constructed kernel matrices on multiple views of expression profiles of primary tumors extracted from pathways. We used Hallmark and Pathway Interaction Database (PID) datasets to restrict the search area to interpretable solutions. We applied our algorithm to 12 cancer cohorts from the Cancer Genome Atlas (TCGA), including more than 5100 primary tumors. The results showed that our algorithm could extract meaningful and disease-specific mechanisms of cancers. We tested the predictive performance of our MLP algorithm and compared it against three existing classification algorithms, namely, random forests, support vector machines, and multiple kernel learning. Our MLP method obtained better or comparable predictive performance against these algorithms.Publication Metadata only Mri-powered magnetic miniature capsule robot with hifu-controlled on-demand drug delivery(Institute of Electrical and Electronics Engineers Inc., 2023) Tiryaki, Mehmet Efe; Dogangun, Fatih; Dayan, Cem Balda; Wrede, Paul; Department of Mechanical Engineering; Sitti, Metin; Department of Mechanical Engineering; College of Engineering; School of MedicineMagnetic resonance imaging (MRI)-guided robotic systems offer great potential for new minimally invasive medical tools, including MRI-powered miniature robots. By re-purposing the imaging hardware of an MRI scanner, the magnetic miniature robot could be navigated into the remote part of the patient's body without needing tethered endoscopic tools. However, state-of-art MRI-powered magnetic miniature robots have limited functionality besides navigation. Here, we propose an MRI-powered magnetic miniature capsule robot benefiting from acoustic streaming forces generated by MRI-guided high-intensity focus ultrasound (HIFU) for controlled drug release. Our design comprises a polymer capsule shell with a submillimeter-diameter drug-release hole that captures an air bubble functioning as a stopper. We use the HIFU pulse to initiate drug release by removing the air bubble once the capsule robot reaches the target location. By controlling acoustic pressure, we also regulate the drug release rate for multiple locations targeting during navigation. We demonstrated that the proposed magnetic capsule robot could travel at high speed, up to 1.13 cm/s in ex vivo porcine small intestine, and release drug to multiple target sites in a single operation, using a combination of MRI-powered actuation and HIFU-controlled release. The proposed MRI-guided microrobotic drug release system will greatly impact minimally invasive medical procedures by allowing on-demand targeted drug delivery.