Meta tags:
description=Map It Anywhere: Empowering BEV Map Prediction using Large-scale Public Dataset;
keywords=MIA, Mapper, Mapillary, OpenStreetMap, BEV, Maps, Semantic Segmentation;

Headings (most frequently used words):

map, mia, data, engine, bev, it, anywhere, prediction, view, predictions, fpv, demo, try, with, the, dataset, in, empowering, using, large, scale, public, datasets, our, empowers, generalizable, bird, eye, from, first, person, images, abstract, different, locations, taking, picture, your, phone, automatically, curates, pairs, worldwide, showing, utility, sampling, wild, untrained, cities, pittsburgh, seattle, zurich, acknowledgments, bibtex, accepted, to, neurips, 2024, and, benchmark, track,

Text of the page (most frequently used words):
and (24), the (20), data (14), for (14), map (13), mia (10), bev (10), engine (8), dataset (8), our (7), #prediction (7), from (6), using (6), this (5), view (5), with (5), fpv (5), code (4), large (4), scale (4), work (4), their (4), that (4), scenarios (4), more (4), demo (4), maps (4), #generalizable (4), source (3), anywhere (3), bird (3), eye (3), public (3), mapillary (3), show (3), locations (3), diverse (3), datasets (3), images (3), you (2), website (2), also (2), your (2), nerfies (2), which (2), empowering (2), mapping (2), cherie (2), zou (2), jiaye (2), alama (2), omar (2), kumar (2), sai (2), mitheran (2), jagadesh (2), chiang (2), benjamin (2), gupta (2), taneesh (2), wang (2), chen (2), keetha (2), nikhil (2), sycara (2), katia (2), scherer (2), sebastian (2), 2024 (2), https (2), arxiv (2), mapitanywhere (2), openstreetmap (2), pittsburgh (2), predictions (2), cities (2), utility (2), sampling (2), different (2), can (2), are (2), have (2), capture (2), automatically (2), try (2), nywhere (2), enables (2), existing (2), platforms (2), camera (2), model (2), promise (2), autonomous (2), navigation (2), first (2), person (2), university (2), use, please, link, back, footer, adapted, template, licensed, under, creative, commons, attribution, sharealike, international, license, inproceedings, ho2024map, title, author, year, booktitle, advances, neural, information, processing, systems, url, org, abs, 2407, 08726, github, com, bibtex, built, incredible, efforts, community, shout, out, paul, edouard, sarlin, his, inspiring, uses, osm, visual, localization, serves, foundation, extend, gratitude, contributors, providing, powers, orienternet, acknowledgments, zurich, seattle, wild, untrained, six, urban, centered, extending, suburbs, selected, highly, populated, new, york, chicago, houston, los, angeles, collect, challenging, dense, traffic, additionally, included, san, francisco, unique, topologies, details, download, links, found, progress, curating, larger, stay, tuned, updates, let, know, suggestions, samples, highlighting, diversity, time, day, seasons, weather, vehicles, pedestrians, page, showing, curates, pairs, worldwide, taking, picture, phone, introduce, seamless, curation, modeling, labeled, open, display, ease, collecting, pair, encompassing, geographies, landscapes, environmental, factors, models, further, train, simple, agnostic, extensive, evaluations, established, benchmarks, curated, zero, shot, performance, far, exceeding, baselines, trained, analysis, highlights, developing, testing, perception, paving, way, robust, effective, pretraining, million, top, down, popular, representation, ground, robot, due, richness, flexibility, downstream, tasks, while, recent, methods, shown, predicting, generalizability, limited, small, regions, captured, current, vehicle, based, context, scalable, approach, towards, enabled, two, crowd, sourced, semantic, abstract, empowers, thread, paper, equal, contribution, buffalo, carnegie, mellon, accepted, neurips, benchmark, track,

Text of the page (random words):
map it anywhere map it anywhere empowering bev map prediction using large scale public datasets accepted to neurips 2024 dataset and benchmark track cherie ho 1 jiaye zou 1 omar alama 1 sai mitheran jagadesh kumar 1 benjamin chiang 1 taneesh gupta 1 chen wang 2 nikhil keetha 1 katia sycara 1 sebastian scherer 1 1 carnegie mellon university 2 university at buffalo equal contribution paper arxiv code data thread data engine demo map prediction demo our m ap i t a nywhere mia data engine empowers generalizable bird s eye view bev map predictions from first person view fpv images abstract 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 m ap i t a nywhere 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 data engine demo try with different locations map prediction demo try taking a picture with your phone mia data engine automatically curates fpv bev pairs worldwide showing mia data engine s utility sampling the mia dataset we show the utility of the mia data engine by sampling six different urban centered locations extending to the suburbs we selected highly populated cities new york chicago houston and los angeles to collect challenging scenarios with diverse and dense traffic additionally we included pittsburgh and san francisco for their unique topologies more details and download links for the dataset can be found in the dataset page we are in the progress of curating a larger dataset with more diverse locations and scenarios stay tuned for updates let us know if you have suggestions samples from the mia dataset highlighting diversity in time of day seasons weather and capture scenarios from vehicles pedestrians in the wild map predictions in untrained cities pittsburgh seattle zurich acknowledgments this work is built on the incredible efforts of the community shout out to paul edouard sarlin and his inspiring work orienternet which uses mapillary and osm for visual localization and serves as the foundation of our work we also extend our gratitude to mapillary openstreetmap and their contributors for providing the data that powers our work bibtex inproceedings ho2024map title map it anywhere mia empowering bird s eye view mapping using large scale public data author ho cherie and zou jiaye and alama omar and kumar sai mitheran jagadesh and chiang benjamin and gupta taneesh and wang chen and keetha nikhil and sycara katia and scherer sebastian year 2024 booktitle advances in neural information processing systems url https arxiv org abs 2407 08726 code https github com mapitanywhere mapitanywhere this website is adapted from the nerfies template which is licensed under a creative commons attribution sharealike 4 0 international license if you use the source code of this website please also link back to the nerfies source code in your footer