BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Robotics Institute Carnegie Mellon University - ECPv6.15.12.1//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:Robotics Institute Carnegie Mellon University
X-ORIGINAL-URL:https://www.ri.cmu.edu
X-WR-CALDESC:Events for Robotics Institute Carnegie Mellon University
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-Robots-Tag:noindex
X-PUBLISHED-TTL:PT1H
BEGIN:VTIMEZONE
TZID:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:20240310T070000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:20241103T060000
END:STANDARD
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:20250309T070000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:20251102T060000
END:STANDARD
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:20260308T070000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:20261101T060000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250902T093000
DTEND;TZID=America/New_York:20250902T103000
DTSTAMP:20260922T072753
CREATED:20250827T132807Z
LAST-MODIFIED:20250827T132807Z
UID:148585-1756805400-1756809000@www.ri.cmu.edu
SUMMARY:Object-Centric Grounding for Deployable and Interactive Vision-Language Navigation Agents
DESCRIPTION:Abstract:\nRobots that operate in human-centric environments must integrate perception\, reasoning\, and action across multiple modalities to complete tasks according to user instructions. For these robots\, being able to navigate according to a natural language instruction about the environment is an important capability\, which requires 3D spatial reasoning\, semantic scene understanding\, and the ability to handle vague or ambiguous instructions. Additionally\, the diverse and noisy nature of real-world environments motivates the need for vision-language navigation (VLN) systems that are robust and able to adaptively generalize. This thesis makes two main contributions toward robust\, interactive vision-language robotic systems by focusing on the underlying task of 3D object-centric grounding. First\, we introduce IRef-VLA\, a large-scale 3D benchmark with millions of referential statements and semantic relations\, including imperfect language\, to support the evaluation of models for 3D scene understanding. Second\, we propose SORT3D\, a modular framework for grounding object-referential language in 3D by leveraging large vision and language models\, heuristic spatial reasoning\, and 2D features\, achieving zero-shot generalization to unseen environments and real-time operation on autonomous ground vehicle systems. Furthermore\, we explore future directions for interactive\, dialogue-enabled vision-language navigation by formulating the problem\, exploring existing benchmarks and laying the groundwork for future work in this area. Together\, these contributions advance general navigation systems that are capable of semantic scene understanding and communicating with human users in complex\, real-world settings.\n\nCommittee:\nJi Zhang (co-advisor)\nWenshan Wang (co-advisor)\nYonatan Bisk\nBowen Li
URL:https://www.ri.cmu.edu/event/object-centric-grounding-for-deployable-and-interactive-vision-language-navigation-agents/
LOCATION:Newell-Simon Hall 4305
CATEGORIES:MSR Thesis Presentation,Student Talks
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250911T100000
DTEND;TZID=America/New_York:20250911T113000
DTSTAMP:20260922T072753
CREATED:20250902T182248Z
LAST-MODIFIED:20250902T182248Z
UID:148685-1757584800-1757590200@www.ri.cmu.edu
SUMMARY:Learning for dexterous manipulation with multi-fingered robotic hands
DESCRIPTION:Abstract: \nDexterous robotic manipulation is becoming increasingly crucial as robots transition from factory manufacturing to everyday human environments such as household assistance and healthcare support. However\, operating in these unstructured settings presents significant challenges in both hardware and software. Achieving adaptability across diverse tasks requires multi-fingered robotic hands with high degrees of freedom\, capable of dexterous manipulation\, together with responsive\, autonomous control policies for precise and robust skill execution. This thesis explores integrated systems for dexterous manipulation using multi-fingered robotic hands. \nWe first introduce DeltaHands\, a modular dexterous hand framework based on Delta robots. DeltaHands are highly dexterous and are simple to fabricate using low-cost\, off-the-shelf materials. Their modular Delta fingers enable flexible design configurations for different applications\, while their parallel and translational kinematics simplify control despite the high degrees of freedom. This framework offers a broad hand design space for dexterous manipulation. \nBuilding on DeltaHands\, we develop Tilde\, an imitation-learning-based in-hand manipulation system. To collect demonstrations\, we propose two teleoperation methods: (1) a vision-based human hand motion tracking interface and (2) a kinematic twin for direct control. Using these demonstrations\, we train control policies with diffusion-based imitation learning. We show that the learned policies can be deployed on a real-world DeltaHand to perform a variety of in-hand dexterous manipulation tasks. \nTo further reduce data collection effort and improve policy generalization\, we introduce ExoStart\, a real-to-sim-to-real learning pipeline that leverages sensorized exoskeleton demonstrations. By capturing direct human–object interaction without robots in the loop\, we use these demonstrations to bootstrap a simulation-based auto-curriculum reinforcement learning method and then transfer the learned policies to real-world robots in zero-shot. Our approach requires fewer than 15 human demonstrations and relies only on sparse reward design\, yet enables the learning of diverse and highly dexterous real-world robotic hand behaviors. \nFinally\, we propose two future directions: (1) enhancing fine-grained manipulation by integrating tactile sensors into fingertips and incorporating multi-sensing modalities into policy learning; (2) exploring hand structure design to enable whole-hand manipulation to improve the grasp stability and strength; and control strategies to transit dexterous fingertip manipulation to powerful grasp. \nThesis Committee Members: \nZeynep Temel (co-chair) \nOliver Kroemer (co-chair) \nNancy Pollard \nOliver Brock (TU Berlin) \nThesis proposal document draft
URL:https://www.ri.cmu.edu/event/learning-for-dexterous-manipulation-with-multi-fingered-robotic-hands/
LOCATION:NSH 4305
CATEGORIES:PhD Thesis Proposal,Student Talks
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250912T100000
DTEND;TZID=America/New_York:20250912T120000
DTSTAMP:20260922T072753
CREATED:20250903T174634Z
LAST-MODIFIED:20250903T174652Z
UID:148739-1757671200-1757678400@www.ri.cmu.edu
SUMMARY:Designing Supportive Agents using the Lived Experience as a Framework for Participatory and Community-based Design to Enhance the Assistive Technology Ecosystem
DESCRIPTION:Abstract\n\n\nIn assistive robotics – a sub-discipline within the broader field of human-robot interaction (HRI) concerned with designing autonomous agents that can support people with disabilities (PWDs) – HRI designers use User-Centered Design (UCD) approaches to create assistive robots and technologies. They involve potential end-users with disabilities as a way of ensuring the end-product\, that is\, the assistive robot\, is usable; in this process\, the focus remains on creating an assistive robot that is predicated on the professional designer’s (i.e.\, the roboticists) understanding of the problem space and evaluating the system’s performance. While Participatory Design (PD) paradigms – democratic approaches aimed at shifting power relations and integrating community interests in design processes outcomes  – are becoming commonplace in HRI\, designers seldom interrogate how or whether these systems fit into the daily lives and address the needs of PwDs; furthermore\, PD or cooperative design (“co-design”) engagements facilitated by HRI researchers often include HRI professionals whose technical expertise is highly regarded\, while the involvement of PwDs is relegated to that as the “expert user” used to sanity check design ideations. Still\, designing with non-professional designers\, including PwDs\, can be challenging. People who are unfamiliar with the process of design\, such as design thinking\, which is the practice of generating and implementing design ideations\, are often not versed with the jargon and methods that comprise professional design practice\, which can be intimidating and exclusionary. \nIn this thesis\, I present the lived experience as a framework for co-designing physically and socially supportive agents (i.e.\, the behaviors\, functions\, and features) with people with disabilities (PwDs) who are non-professional designers. Throughout this thesis\, I demonstrate robotic technologies that are functionally “assistive” – that is\, they assist in complex daily tasks\, such as robotic navigation devices; I also present robotic technologies that are “supportive” (which comprises “assistive” agents); that is\, they connect people to resources\, such as delivery robots\, and activities\, such as social agents to encourage or maintain emotional wellbeing. Through this work\, I reveal that situating design ideation in non-professional designers’ lived experiences (i.e.\, the tasks they perform daily\, the challenges that emerge\, and the ways that they envision assistive and supportive robotic technologies intervening in and remediating these challenges) yields rich and specific design recommendations. Additionally\, our findings suggest that participatory encounters with PwDs that are grounded in the lived experience can be made more approachable and sustain participation in iterative design engagements (as evidenced through turnout in multi-session encounters). \nIn this thesis\, I provide theoretical grounding for the “lived experience” across philosophical and critical study\, technology design practice\, and robotics and present three works (1 completed\, 2 in-progress) that demonstrate the effectiveness of engaging with the lived experience of non-professional designers with disability. These works include: 1) co-designing robotic mobility devices with who are blind or have low vision; 2) pluriversal (i.e.\, the richness and diversity of lived experience that comprises a single person’s life) design engagements with people with mobility and visual disabilities on improving the accessibility of on-demand\, last-mile delivery delivery robots; and finally\, 3) first-person and co-design engagements towards designing socially assistive agents that support emotional wellbeing by facilitating joint artmaking and co-regulation strategies. Through this work\, I provide design recommendations for developers creating robotics technologies that are aimed at supporting PwDs. Furthermore\, I demonstrate that PD practices that hinge on the lived experience bolster engagement and yield rich insights that can inform specific design recommendations for assistive robots. \n\n\nThesis Committee Members:\nAaron Steinfeld\, Chair\nJean Oh\, RI\nPatrick Carrington\, HCII\nCynthia Bennett\, Google
URL:https://www.ri.cmu.edu/event/148739/
LOCATION:GHC 6501
CATEGORIES:PhD Thesis Proposal,Student Talks
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250912T160000
DTEND;TZID=America/New_York:20250912T170000
DTSTAMP:20260922T072753
CREATED:20250909T144635Z
LAST-MODIFIED:20250909T144635Z
UID:148791-1757692800-1757696400@www.ri.cmu.edu
SUMMARY:Design and Validation of Minimally-Actuated Mobility for Planetary Rovers
DESCRIPTION:Abstract: \nThis thesis investigates low degree-of-freedom (DoF)\, minimally-actuated rover mobility architectures for planetary exploration. These platforms seek to reduce mass and complexity compared to the heritage rocker-bogie system while maintaining functional performance. The work focuses on two systems: MoonRanger\, a lightweight lunar micro-rover slated for a 2029 moon mission\, and Zoë2\, a research rover developed as a successor to CMU’s 22-year old Zoë1 rover. \nFor MoonRanger\, we analyze the limitations of the gravity offload method for Earth-based mobility testing and validate alternative approaches using granular scaling laws\, single-wheel experiments\, and high-fidelity discrete element modeling (DEM). Results demonstrate that offload testing significantly overestimates performance\, while full-mass testing and discrete-element modeling provide more accurate predictions of lunar behavior. A wheel design study using black-box optimization is initiated to identify promising configurations for future development. \nThe second part of this thesis details the design\, implementation\, and evaluation of Zoë2\, a passive-steering rover optimized for energy efficiency and maneuverability. Experimental validation confirms that the redesigned mechanical\, electrical\, and software systems provide a versatile testbed for planetary autonomy research. We find that compared to skid steering\, passive steering reduces power consumption by up to 30% while simultaneously allowing far more precise blind navigation. \n\nTogether\, these studies highlight the tradeoffs between simplicity\, efficiency\, and mobility performance in planetary rover design. The findings contribute to the development of lighter\, more capable mobility systems for future lunar and planetary missions. \n  \nCommittee: \nDavid Wettergreen (advisor) \nRed Whittaker \nDan McGann
URL:https://www.ri.cmu.edu/event/design-and-validation-of-minimally-actuated-mobility-for-planetary-rovers/
LOCATION:GHC 4405
CATEGORIES:MSR Thesis Presentation,Student Talks
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250916T140000
DTEND;TZID=America/New_York:20250916T160000
DTSTAMP:20260922T072753
CREATED:20250909T133057Z
LAST-MODIFIED:20250909T133057Z
UID:148788-1758031200-1758038400@www.ri.cmu.edu
SUMMARY:Getting Optimization layers to play well with Deep Networks : Numerical methods and Architectures
DESCRIPTION:Abstract: \nMany real-world challenges\, from robotic control to resource management\, can be effectively formulated as optimization problems. Recent advancements have focused on incorporating these optimization problems as layers within deep learning pipelines\, enabling the explicit inclusion of auxiliary constraints or cost functions\, which is crucial for applications such as enforcing physical laws\, ensuring safety constraints\, and optimizing complex objectives. However\, these layers introduce several challenges\, including inference inefficiencies\, unstable training dynamics\, modeling inaccuracies\, and representational inefficiencies\, which need to be addressed to fully harness their potential. \nWe systematically investigate these challenges and propose novel numerical methods and architectural solutions that mitigate them\, making optimization layers more efficient and effective within deep learning pipelines. Our contributions include methods for enhancing computational efficiency by exploiting the iterative nature of optimization problems\, tackling issues of gradient bias and variance in high dimensional problems by exploiting parallelism and network learnt priors about the system\, improving sample efficiency in reinforcement learning using approximate simulators\, and mitigating representational problems with using complicated constrained optimization layers by creating a tight feedback loop between the optimizer state and the network outputs in domains like robotic control and mechanism design with LLMs. We demonstrate these contributions across different applications\, ranging from input-optimization problems\, 3D pose estimation and reconstruction\, differentiable model predictive control and reinforcement learning problems. We also present a new approach for visual-inertial navigation in nanosatellites\, highlighting the practical benefits of integrating optimization layers in challenging real-world scenarios. \n\nTogether\, these contributions advance our understanding of the complexities and opportunities in integrating optimization layers within deep learning models\, offering new frameworks and insights that improve efficiency\, stability\, and generalizability across a wide range of complex tasks.\n\nThesis Committee Members:\nZico Kolter\, Co-chair\nZac Manchester\, Co-chair\nGeoffrey Gordon\nMax Simchowitz\nVladlen Koltun\, Apple
URL:https://www.ri.cmu.edu/event/getting-optimization-layers-to-play-well-with-deep-networks-numerical-methods-and-architectures-2/
LOCATION:Newell-Simon Hall 4305
CATEGORIES:PhD Thesis Defense,Student Talks
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250918T140000
DTEND;TZID=America/New_York:20250918T163000
DTSTAMP:20260922T072753
CREATED:20250903T175819Z
LAST-MODIFIED:20250910T173711Z
UID:148750-1758204000-1758213000@www.ri.cmu.edu
SUMMARY:Scaling Laws Revisited: When Data\, Not Compute\, is the Bottleneck
DESCRIPTION:Abstract:\nThe formula for AI progress has long appeared straightforward: more compute\, more data\, better models. Yet while compute is growing with better hardware and bigger clusters\, data is stagnating—calling into question the very scaling laws that powered the last decade. The internet—often described as the “fossil fuel” of AI—offers only a finite reservoir of training data\, raising a critical question: how can we sustain the scaling trends that underpin modern AI progress?\n\n\n\n\n\n\n\nThis thesis explores one promising direction: trading off compute for data. The central idea is to leverage additional compute to compensate for limited data. We present three simple strategies towards this goal: \n\nMake the task harder. We show that making the training objective more challenging can improve the generalization ability of current models.\nMake the supervision richer. We demonstrate that providing dense gradient feedback can enhance the sample-efficiency for post-training foundation models.\n\n\nMake the tasks unsupervised. We find that large language models can improve at reasoning without access to any ground truth question–answer pairs\, reducing reliance on costly supervision.\n\nCollectively\, the results suggest new ways for extending .scaling laws in an era where data growth can no longer be taken for granted\n\n\n\n\n\n\n\n\n\nThesis Committee Members: \nDeepak Pathak\, Co-Chair \nKaterina Fragkiadak\, Co-Chair \nDeva Ramanan \nHao Liu \nYejin Choi\, Stanford University
URL:https://www.ri.cmu.edu/event/mihir-prabhudesai/
LOCATION:NSH 3305
CATEGORIES:PhD Thesis Proposal,Student Talks
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250919T100000
DTEND;TZID=America/New_York:20250919T120000
DTSTAMP:20260922T072753
CREATED:20250911T175647Z
LAST-MODIFIED:20250911T175647Z
UID:148841-1758276000-1758283200@www.ri.cmu.edu
SUMMARY:Accessible Dexterous Manipulation with Soft Hands: Designs\, Methods\, Models\, and the DexKit Platform
DESCRIPTION:Abstract: \nRobot dexterity remains an open challenge in robotics that has the potential to transform\nmanufacturing\, healthcare\, and daily life. Robots that safely and robustly interact with unstructured\nenvironments must combine compliant hardware with models and planners that tolerate uncertainty.\nAdditionally\, if robust robot dexterity is to be realized outside of research labs\, it must be accessible\nto a broad audience\, with low-cost hardware and open-source software. \nThis thesis advances dexterous manipulation with soft\, tendon-driven hands by integrating new\nfabrication and control methods\, data-driven models of manipulation capabilities\, a fast algorithm\nfor robust grasp synthesis\, and a first-of-its-kind accessible experimental platform. \nI introduce fully soft foam hands actuated by tendons routed on textile skins. I detail a\nsimple molding-and-casting pipeline\, validate a soft-body simulation framework\, compare inverse-\nkinematics control strategies\, and optimize nontrivial tendon routings. I further report a user study\non human-designed routings\, demonstrations of power/precision grasps and in-hand manipulation\,\nsub-millimeter repeatability\, and year-long durability\, alongside an analysis of limitations (e.g.\,\nrouting through foam\, sensing\, and sim-to-real gaps). \nTo tackle the challenge of planning with soft hands\, I develop data-driven models of manipulation\ncapabilities that capture the inherent uncertainty and redundancy of soft hands. Additionally I\ndemonstrate a fast\, anytime method to compute globally optimal Independent Contact Regions\n(ICRs) by iteratively building an incremental Delaunay triangulation over grasp configuration space.\nI show that ICRs guide simple policies that remain robust to real-world uncertainties in object size\,\npose\, and geometry. \nFinally\, to promote accessibility\, I contribute the DexKit system\, a low-cost\, anthropomorphic system (12\nactuated DoF hand on a 4-DoF gantry) that can be built for under $2000. \nTheses Committee Members:\nNancy Pollard (chair)\nMatthew Mason\nChristopher Atkeson\nJames Bern (Williams College)\n\nDraft of the Thesis Proposal Document Link
URL:https://www.ri.cmu.edu/event/accessible-dexterous-manipulation-with-soft-hands-designs-methods-models-and-the-dexkit-platform/
LOCATION:GHC 4405
CATEGORIES:PhD Thesis Proposal,Student Talks
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250922T130000
DTEND;TZID=America/New_York:20250922T150000
DTSTAMP:20260922T072753
CREATED:20250903T175414Z
LAST-MODIFIED:20250903T175414Z
UID:148748-1758546000-1758553200@www.ri.cmu.edu
SUMMARY:Robust Incremental Distributed Collaborative Simultaneous Localization and Mapping
DESCRIPTION:Abstract:\nMulti-robot teams show exceptional promise across applications like Search-and-Rescue\, disaster-response\, agriculture\, forestry\, and scientific exploration due to their ability to go where humans cannot\, parallelize activity\, operate robustly to failures\, and expand capabilities beyond that of an individual robot. Collaborative Simultaneous Localization and Mapping (C-SLAM) is a fundamental capability for these multi-robot teams as it is required for them to plan\, navigate\, and\, in turn\, achieve their mission goals. A key component of the C-SLAM system is the back-end algorithm responsible for estimating the state of the robot team from their distributed\, noisy measurements. However\, existing C-SLAM back-end algorithms struggle to handle the practical conditions experienced by multi-robot teams deployed in the real-world. During real-world deployments multi-robot teams require C-SLAM back-end algorithms that are — 1) online to update estimates as new information is gathered\, 2) robust to outlier data that we expect due to perceptual aliasing\, 3) resilient to sparse and unreliable communication networks\, 4) provide accurate and consistent solutions to the robot team\, and 5) are scalable to large multi-robot teams we expect in future applications. In this thesis we propose a C-SLAM back-end algorithm that achieves all of these goals. We first explore Consensus Alternating Direction Method of Multipliers (C-ADMM) as a theoretical basis for C-SLAM algorithms. We then explore how to extend this base method to address the practicalities of real-world operation. Along the way we develop a novel incremental SLAM algorithm that addresses outlier measurements incrementally and in real-time. We complete these extensions with riMESA — a novel robust\, incremental\, distributed C-SLAM back-end algorithm designed for real-world C-SLAM problems. Finally\, we develop a suite of benchmark C-SLAM datasets based on real-world data that are used to test and validate the proposed algorithm.\n\n\nThesis Committee Members: \nMichael Kaess\, Chair \nSebastian Scherer\n \nGeorge Kantor \nTim Barfoot\, University of Toronto
URL:https://www.ri.cmu.edu/event/robust-incremental-distributed-collaborative-simultaneous-localization-and-mapping-2/
LOCATION:NSH 4305
CATEGORIES:PhD Thesis Defense,Student Talks
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250923T120000
DTEND;TZID=America/New_York:20250923T133000
DTSTAMP:20260922T072753
CREATED:20250915T140520Z
LAST-MODIFIED:20250915T140520Z
UID:148847-1758628800-1758634200@www.ri.cmu.edu
SUMMARY:Towards 4D perception with foundational priors
DESCRIPTION:Abstract:\nAs humans\, we are constantly interacting with and observing a three-dimensional dynamic world. Building this spatiotemporal or 4D understanding in vision algorithms is not straightforward as there is orders of magnitude less 4D data than 2D images and videos. This underscores the need to find meaningful ways to exploit 2D data to realize 4D tasks. Recent advancements in building “foundation models” — that have learnt generative/structural priors in a data-driven manner from internet-scale data — have allowed us access to these rich real-world priors for free. In this thesis\, we investigate how one can tune these priors for 4D perception tasks like amodal tracking and completion\, dynamic reconstruction and next-timestep prediction. \nWe pursue three complementary directions. First\, in the absence of foundational priors\, we build these ourselves in a self-supervised manner via the task of next-timestep prediction using sequences of 3D LiDAR sweeps of dynamic scenes. Importantly\, we show that bottlenecking next-timestep prediction with a 4D representation is crucial. We find that such a forecasting model can be used for downstream motion planning for autonomous vehicles\, which helps reduce collision rates to a large extent. \nSecond\, we capitalize on foundational priors in a zero-shot manner. We turn to large reconstruction models that predict per pixel depth for images and videos. We use these to solve two underconstrained tasks — (1) tracking objects across occlusions in 2.5D\, and (2) reconstructing dynamic scenes from sparse-views. In both settings\, we find that one can do drastically better than prior state-of-the-art using additional scene cues in the form of data-driven depth priors. \nThird\, we exploit foundational priors via finetuning. We specifically look at video diffusion models and reformulate amodal perception and dynamic novel-view synthesis into self-supervised tasks that video models are good at i.e. inpainting. We find that it is surprisingly light-weight\, in terms of data and compute\, to finetune video diffusion models. This suggests that concepts similar to human visual perception are embedded in foundation models\, which only have to be “controlled” to perform other tasks. \nTogether these contributions highlight how one can build\, leverage and adapt foundational priors for spatiotemporal perception in a scalable manner — the scale is enabled by relying increasingly on internet-scale 2D data and carefully designing self-supervised objectives for learning. \n\n\nThesis Committee Members: \nDeva Ramanan\, Chair \nShubham Tulsiani\n \nKaterina Fragkiadaki \nCarl Vondrick\, Columbia \nLeonidas Guibas\, Stanford & Google \nLink to thesis draft
URL:https://www.ri.cmu.edu/event/towards-4d-perception-with-foundational-priors/
LOCATION:Newell-Simon Hall 3305
CATEGORIES:PhD Thesis Defense,Student Talks
END:VEVENT
END:VCALENDAR