Look for the Change: Learning Object States and State-Modifying Actions from Untrimmed Web Videos

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21730%2F22%3A00358883" target="_blank" >RIV/68407700:21730/22:00358883 - isvavai.cz</a>

Výsledek na webu

<a href="https://doi.org/10.1109/CVPR52688.2022.01357" target="_blank" >https://doi.org/10.1109/CVPR52688.2022.01357</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1109/CVPR52688.2022.01357" target="_blank" >10.1109/CVPR52688.2022.01357</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Look for the Change: Learning Object States and State-Modifying Actions from Untrimmed Web Videos

Popis výsledku v původním jazyce

Human actions often induce changes of object states such as “cutting an apple”, “cleaning shoes” or “pouring coffee”. In this paper, we seek to temporally localize ob ject states (e.g. “empty” and “full” cup) together with the corresponding state-modifying actions (“pouring coffee”) in long uncurated videos with minimal supervision. The contributions of this work are threefold. First, we develop a self-supervised model for jointly learning state-modifying actions together with the corresponding object states from an uncurated set of videos from the Internet. The model is self-supervised by the causal ordering signal, i.e. initial ob ject state -> manipulating action -> end state. Second, to cope with noisy uncurated training data, our model incor porates a noise adaptive weighting module supervised by a small number of annotated still images, that allows to ef ficiently filter out irrelevant videos during training. Third, we collect a new dataset with more than 2600 hours of video and 34 thousand changes of object states, and manually an notate a part of this data to validate our approach. Our re sults demonstrate substantial improvements over prior work in both action and object state-recognition in video.

Název v anglickém jazyce

Look for the Change: Learning Object States and State-Modifying Actions from Untrimmed Web Videos

Popis výsledku anglicky

Human actions often induce changes of object states such as “cutting an apple”, “cleaning shoes” or “pouring coffee”. In this paper, we seek to temporally localize ob ject states (e.g. “empty” and “full” cup) together with the corresponding state-modifying actions (“pouring coffee”) in long uncurated videos with minimal supervision. The contributions of this work are threefold. First, we develop a self-supervised model for jointly learning state-modifying actions together with the corresponding object states from an uncurated set of videos from the Internet. The model is self-supervised by the causal ordering signal, i.e. initial ob ject state -> manipulating action -> end state. Second, to cope with noisy uncurated training data, our model incor porates a noise adaptive weighting module supervised by a small number of annotated still images, that allows to ef ficiently filter out irrelevant videos during training. Third, we collect a new dataset with more than 2600 hours of video and 34 thousand changes of object states, and manually an notate a part of this data to validate our approach. Our re sults demonstrate substantial improvements over prior work in both action and object state-recognition in video.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

<a href="/cs/project/EF15_003%2F0000468" target="_blank" >EF15_003/0000468: Inteligentní strojové vnímání</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2022

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Proceeding 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

ISBN

978-1-6654-6946-3

ISSN

1063-6919

e-ISSN

2575-7075

Počet stran výsledku

11

Strana od-do

13936-13946

Název nakladatele

IEEE

Místo vydání

Piscataway

Místo konání akce

New Orleans, Louisiana

Datum konání akce

19. 6. 2022

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

000870759107004