Meta tags:
author=Nikhil Varma Keetha;
description=Personal website for Nikhil Varma Keetha. ;
keywords=nikhil, varma, keetha, robotics, computer-vision, ai, academic-website, portfolio-website;

Headings (most frequently used words):

nikhil, varma, keetha, news, featured, publications, 2024, 2023, 2022, 2021,

Text of the page (most frequently used words):
and (49), the (44), for (34), 2021 (17), 2022 (15), that (13), 2023 (13), robotics (13), map (12), cmu (12), object (11), from (11), our (10), 2024 (9), with (9), vpr (9), utility (9), place (8), performance (8), using (8), #nikhil (7), keetha (7), have (7), this (7), ieee (7), bev (7), visual (6), approaches (6), specific (6), environments (6), are (6), real (6), data (6), mar (6), while (5), code (5), pdf (5), abs (5), iros (5), identification (5), airobject (5), scale (5), sebastian (5), scherer (5), rendering (5), show (5), maps (5), mia (5), reviewed (5), live (5), recognition (4), high (4), can (4), novel (4), approach (4), two (4), environment (4), these (4), state (4), art (4), datasets (4), time (4), semantic (4), across (4), video (4), encoding (4), autonomous (4), scene (4), understanding (4), localization (4), representation (4), robot (4), based (4), cvpr (4), anyloc (4), existing (4), may (4), anywhere (4), krishna (4), dense (4), mapping (4), slam (4), gaussians (4), camera (4), splatam (4), more (4), view (4), navigation (4), generalizable (4), prediction (4), large (4), institute (4), sep (4), now (4), varma (3), descriptors (3), unsupervised (3), which (3), feature (3), achieves (3), further (3), analysis (3), how (3), sourav (3), garg (3), model (3), such (3), objects (3), systems (3), single (3), global (3), multiple (3), world (3), temporal (3), graph (3), robust (3), first (3), methods (3), chen (3), wang (3), structured (3), their (3), universal (3), models (3), enables (3), website (3), thread (3), murthy (3), field (3), deva (3), ramanan (3), fpv (3), current (3), public (3), perception (3), mentored (3), prof (3), aug (3), 2020 (3), received (3), nov (3), intern (3), summer (3), series (3), accepted (3), phd (3), feb (3), wild (3), research (3), meta (3), best (2), way (2), via (2), attempted (2), places (2), identifying (2), cues (2), regions (2), range (2), employ (2), learning (2), keypoint (2), selection (2), combining (2), measures (2), demonstrating (2), results (2), categorization (2), has (2), higher (2), than (2), building (2), michael (2), milford (2), vital (2), tasks (2), exploration (2), track (2), limited (2), viewpoint (2), temporally (2), evolving (2), unknown (2), agnostic (2), context (2), obtain (2), embeddings (2), specifically (2), method (2), demonstrate (2), blog (2), most (2), urban (2), unstructured (2), deployment (2), work (2), solution (2), works (2), training (2), representations (2), derived (2), self (2), supervised (2), build (2), features (2), experiments (2), foundation (2), jatavallabhula (2), jay (2), karhade (2), towards (2), recent (2), shown (2), enable (2), reconstruction (2), rgb (2), optimization (2), tracking (2), extensive (2), bird (2), eye (2), promise (2), images (2), platforms (2), engine (2), dataset (2), developing (2), publications (2), interning (2), qut (2), centre (2), mila (2), scholar (2), automation (2), letters (2), airlab (2), apr (2), tartan (2), heaputil (2), jun (2), presented (2), part (2), riss (2), icra (2), oct (2), offer (2), robolaunch (2), dec (2), master (2), working (2), pittsburgh (2), visiting (2), researcher (2), degree (2), also (2), copyright, powered, theme, last, updated, october, folio, jekyll, reach, email, typically, match, image, landmarks, but, concept, not, singular, rather, take, forms, paper, present, deduce, key, types, particular, contrastive, principles, estimate, both, vector, locally, aggregated, vlad, clusters, manner, then, used, guide, local, matching, through, three, challenging, benchmark, simultaneously, reducing, required, storage, compute, provide, cluster, semantically, meaningful, finer, grained, often, level, road, characterise, vary, different, hierarchical, dual, robotic, previous, either, generate, however, fixed, partial, setup, there, requirement, built, observes, viewpoints, furthermore, given, vast, distribution, process, must, class, propose, dubbed, generated, convolutional, network, structural, information, frames, obtained, attention, severe, occlusion, perceptual, aliasing, shift, deformation, transform, outperforming, frame, sequential, knowledge, one, poster, kuan, yuheng, qiu, embedding, date, performant, task, they, exhibit, strong, predominantly, driving, degrades, severely, brittle, develop, technique, broad, outdoors, indoors, aerial, underwater, subterranean, without, any, fine, tuning, general, purpose, off, shelf, right, substrate, upon, aggregation, suite, achieve, significantly, improvement, characterizing, properties, uncovering, unique, domains, encapsulate, similar, detailed, lay, solutions, deployed, anytime, anyview, encourage, readers, explore, project, page, interactive, demos, https, github, madhava, avneesh, mishra, simultaneous, pivotal, embodied, quality, scenes, posed, cameras, light, representing, unposed, monocular, addresses, limitations, prior, radiance, including, fast, ability, determine, areas, been, previously, mapped, expansion, adding, online, pipeline, tailoring, use, underlying, gaussian, silhouette, guided, differentiable, pose, estimation, construction, synthesis, its, superiority, over, allowing, resolution, jonathon, luiten, gengshan, yang, splat, top, down, popular, ground, due, richness, flexibility, downstream, predicting, person, generalizability, small, captured, vehicle, scalable, enabled, crowd, sourced, mapillary, openstreetmap, introduce, seamless, curation, modeling, labeled, open, source, display, ease, automatically, collecting, million, pair, encompassing, diverse, geographies, landscapes, environmental, factors, capture, scenarios, train, simple, evaluations, established, benchmarks, curated, effective, pretraining, zero, shot, far, exceeding, baselines, trained, highlights, testing, paving, neurips, jiaye, zou, omar, alama, sai, mitheran, jagadesh, kumar, benjamin, chiang, taneesh, gupta, authors, cherie, empowering, featured, accv, diversity, grant, winter, liam, paull, jan, selected, organizing, community, seminar, experience, fully, funded, eccv, graduated, tech, distinction, organized, initiative, reached, 000, students, viewers, worldwide, reviewing, fall, conceptfusion, planning, collaborating, dynamic, gps, denied, drone, heli, survey, defended, thesis, scientist, zurich, peter, kontschieder, jul, continuing, news, apart, huge, movie, buff, find, recreation, cooking, music, anime, gaming, bachelor, engineering, physics, during, undergrad, had, pleasure, was, grateful, many, wonderful, mentors, collaborators, iit, ism, dhanbad, india, focuses, spatial, robots, interests, lie, intersection, geometric, computer, vision, semi, innate, passion, human, deployable, positive, societal, impact, hello, student, carnegie, mellon, university, advised, currently, team, 15232, 1723, murray, ave, sqh, wander, into, service, outreach, projects, about, toggle,

Text of the page (random words):
nikhil varma keetha toggle navigation about current blog publications projects service outreach cv nikhil varma keetha phd in robotics the robotics institute cmu wander into the unknown ️ ri sqh cmu 1723 murray ave pittsburgh pa 15232 hello i m a phd student at carnegie mellon university cmu advised by sebastian scherer deva ramanan as a part of the airlab i m also currently a visiting researcher at meta working with the xr maps team my current research focuses on developing generalizable spatial perception systems for scene understanding and deployment of field robots in the wild my interests lie at the intersection of geometric computer vision field robotics and large scale semi self supervised learning i have an innate passion for understanding how human perception navigation works and how we can build real world deployable autonomous systems with a positive societal impact i received a master s degree in robotics from cmu and a bachelor s degree in engineering physics from iit ism dhanbad india during my undergrad i had the pleasure to intern at mila and qut centre for robotics and was also a 2021 cmu robotics institute summer scholar i m grateful to have many wonderful mentors and collaborators apart from research i m a huge movie buff and find recreation in cooking music anime and gaming news sep 1 2024 continuing at meta pittsburgh as a visiting researcher jul 19 2024 map it anywhere mia is now live may 20 2024 research scientist intern at meta zurich working with peter kontschieder feb 1 2024 defended my master s thesis at cmu dec 14 2023 our survey on foundation models for robotics is now live dec 1 2023 splatam is now live nov 1 2023 we show gps denied localization on a drone and heli in the wild sep 1 2023 collaborating with deva ramanan on dynamic reconstruction tracking in the wild aug 1 2023 anyloc is live mar 14 2023 tartan planning series is now live feb 14 2023 conceptfusion is public feb 2 2023 received cmu ri phd offer for fall 2023 nov 1 2022 reviewing for cvpr 2023 oct 15 2022 reviewed for ieee icra 2023 sep 1 2022 robolaunch reached 13 000 students viewers worldwide jun 1 2022 organized the robolaunch initiative at cmu may 30 2022 graduated b tech with distinction apr 11 2022 reviewed for eccv 2022 mar 23 2022 reviewed for ieee iros 2022 mar 19 2022 accepted fully funded ms in robotics offer from cmu mar 4 2022 airobject accepted to cvpr 2022 oct 10 2021 reviewed for ieee ra l icra 2022 sep 27 2021 presented heaputil at iros 2021 aug 10 2021 presented riss summer experience as a part of the 2021 riss community seminar series jun 14 2021 heaputil accepted to ieee robotics and automation letters ra l and iros 2021 may 1 2021 organizing the tartan slam series at cmu robotics institute apr 1 2021 interning at the airlab cmu mentored by dr chen wang and prof sebastian scherer mar 25 2021 reviewed for ieee robotics and automation letters ra l iros 2021 mar 18 2021 selected as a cmu robotics institute summer scholar jan 1 2021 winter 2021 intern at real mila mentored by krishna murthy and prof liam paull nov 1 2020 received accv 2020 diversity grant aug 14 2020 interning at the qut centre for robotics mentored by dr sourav garg and prof michael milford featured publications 2024 map it anywhere mia empowering bird s eye view mapping using large scale public data cherie ho jiaye zou omar alama sai mitheran jagadesh kumar benjamin chiang taneesh gupta chen wang and 3 more authors neurips 2024 abs pdf code thread website top down bird s eye view bev maps are a popular representation for ground robot navigation due to their richness and flexibility for downstream tasks while recent methods have shown promise for predicting bev maps from first person view fpv images their generalizability is limited to small regions captured by current autonomous vehicle based datasets in this context we show that a more scalable approach towards generalizable map prediction can be enabled by using two large scale crowd sourced mapping platforms mapillary for fpv images and openstreetmap for bev semantic maps we introduce map it anywhere mia a data engine that enables seamless curation and modeling of labeled map prediction data from existing open source map platforms using our mia data engine we display the ease of automatically collecting a 1 2 million fpv bev pair dataset encompassing diverse geographies landscapes environmental factors camera models capture scenarios we further train a simple camera model agnostic model on this data for bev map prediction extensive evaluations using established benchmarks and our dataset show that the data curated by mia enables effective pretraining for generalizable bev map prediction with zero shot performance far exceeding baselines trained on existing datasets by 35 our analysis highlights the promise of using large scale public maps for developing testing generalizable bev perception paving the way for more robust autonomous navigation splatam splat track map 3d gaussians for dense rgb d slam nikhil keetha jay karhade krishna murthy jatavallabhula gengshan yang sebastian scherer deva ramanan and jonathon luiten cvpr 2024 abs pdf code thread video website dense simultaneous localization and mapping slam is pivotal for embodied scene understanding recent work has shown that 3d gaussians enable high quality reconstruction and real time rendering of scenes using multiple posed cameras in this light we show for the first time that representing a scene by 3d gaussians can enable dense slam using a single unposed monocular rgb d camera our method splatam addresses the limitations of prior radiance field based representations including fast rendering and optimization the ability to determine if areas have been previously mapped and structured map expansion by adding more gaussians we employ an online tracking and mapping pipeline while tailoring it to specifically use an underlying gaussian representation and silhouette guided optimization via differentiable rendering extensive experiments show that splatam achieves up to 2x state of the art performance in camera pose estimation map construction and novel view synthesis demonstrating its superiority over existing approaches while allowing real time rendering of a high resolution dense 3d map 2023 anyloc towards universal visual place recognition nikhil keetha avneesh mishra jay karhade krishna murthy jatavallabhula sebastian scherer madhava krishna and sourav garg ieee ra l 2023 abs pdf code thread video website visual place recognition vpr is vital for robot localization to date the most performant vpr approaches are environment and task specific while they exhibit strong performance in structured environments predominantly urban driving their performance degrades severely in unstructured environments rendering most approaches brittle to robust real world deployment in this work we develop a universal solution to vpr a technique that works across a broad range of structured and unstructured environments urban outdoors indoors aerial underwater and subterranean environments without any re training or fine tuning we demonstrate that general purpose feature representations derived from off the shelf self supervised models with no vpr specific training are the right substrate upon which to build such a universal vpr solution combining these derived features with unsupervised feature aggregation enables our suite of methods anyloc to achieve up to 4x significantly higher performance than existing approaches we further obtain a 6 improvement in performance by characterizing the semantic properties of these features uncovering unique domains which encapsulate datasets from similar environments our detailed experiments and analysis lay a foundation for building vpr solutions that may be deployed anywhere anytime and across anyview we encourage the readers to explore our project page and interactive demos https anyloc github io 2022 airobject a temporally evolving graph embedding for object identification nikhil keetha chen wang yuheng qiu kuan xu and sebastian scherer cvpr 2022 abs pdf blog code poster object encoding and identification are vital for robotic tasks such as autonomous exploration semantic scene understanding and re localization previous approaches have attempted to either track objects or generate descriptors for object identification however such systems are limited to a fixed partial object representation from a single viewpoint in a robot exploration setup there is a requirement for a temporally evolving global object representation built as the robot observes the object from multiple viewpoints furthermore given the vast distribution of unknown novel objects in the real world the object identification process must be class agnostic in this context we propose a novel temporal 3d object encoding approach dubbed airobject to obtain global keypoint graph based embeddings of objects specifically the global 3d object embeddings are generated using a temporal convolutional network across structural information of multiple frames obtained from a graph attention based encoding method we demonstrate that airobject achieves the state of the art performance for video object identification and is robust to severe occlusion perceptual aliasing viewpoint shift deformation and scale transform outperforming the state of the art single frame and sequential descriptors to the best of our knowledge airobject is one of the first temporal object encoding methods 2021 a hierarchical dual model of environment and place specific utility for visual place recognition nikhil keetha michael milford and sourav garg ieee ra l iros 2021 abs pdf code video visual place recognition vpr approaches have typically attempted to match places by identifying visual cues image regions or landmarks that have high utility in identifying a specific place but this concept of utility is not singular rather it can take a range of forms in this paper we present a novel approach to deduce two key types of utility for vpr the utility of visual cues specific to an environment and to a particular place we employ contrastive learning principles to estimate both the environment and place specific utility of vector of locally aggregated descriptors vlad clusters in an unsupervised manner which is then used to guide local feature matching through keypoint selection by combining these two utility measures our approach achieves state of the art performance on three challenging benchmark datasets while simultaneously reducing the required storage and compute time we provide further analysis demonstrating that unsupervised cluster selection results in semantically meaningful results that finer grained categorization often has higher utility for vpr than high level semantic categorization e g building road and characterise how these two utility measures vary across different places and environments the best way to reach me is via email copyright 2024 nikhil varma keetha powered by jekyll with al folio theme last updated october 23 2024