HTTP/2 301
server: GitHub.com
content-type: text/html
location: https://joaanna.github.io/something_else/
x-github-request-id: FC05:123DE:7D4ECD:8C87BA:695165C3
accept-ranges: bytes
age: 0
date: Sun, 28 Dec 2025 17:15:56 GMT
via: 1.1 varnish
x-served-by: cache-bom-vanm7210082-BOM
x-cache: MISS
x-cache-hits: 0
x-timer: S1766942156.094567,VS0,VE198
vary: Accept-Encoding
x-fastly-request-id: bfc7f21e80ddb9050437dbcfd3532b22517692dc
content-length: 162
HTTP/2 200
server: GitHub.com
content-type: text/html; charset=utf-8
last-modified: Fri, 25 Aug 2023 00:43:43 GMT
access-control-allow-origin: *
etag: W/"64e7f93f-3ed9"
expires: Sun, 28 Dec 2025 17:25:56 GMT
cache-control: max-age=600
content-encoding: gzip
x-proxy-cache: MISS
x-github-request-id: 35EE:2C10E1:7DC03C:8CF8D4:695165CC
accept-ranges: bytes
age: 0
date: Sun, 28 Dec 2025 17:15:56 GMT
via: 1.1 varnish
x-served-by: cache-bom-vanm7210082-BOM
x-cache: MISS
x-cache-hits: 0
x-timer: S1766942156.306319,VS0,VE213
vary: Accept-Encoding
x-fastly-request-id: 5e7b3222f2f292c580765ee00423404f0cd5293a
content-length: 3937
SomethingElse
Something-Else: Compositional Action Recognition with Spatial-Temporal Interaction Networks
CVPR 2020
Human action is naturally compositional: humans can easily recognize and perform actions with objects that are different from those used in training demonstrations. In this paper, we study the compositionality of action by looking into the dynamics of subject-object interactions. We propose a novel model which can explicitly reason about the geometric relations between constituent objects and an agent performing an action. To train our model, we collect dense object box annotations on the Something-Something dataset. We propose a novel compositional action recognition task where the training combinations of verbs and nouns do not overlap with the test set. The novel aspects of our model are applicable to activities with prominent object interaction dynamics and to objects which can be tracked using state-of-the-art approaches; for activities without clearly defined spatial object-agent interactions, we rely on baseline scene-level spatio-temporal representations. We show the effectiveness of our approach not only on the proposed compositional action recognition task, but also in a few-shot compositional setting which requires the model to generalize across both object appearance and action category.
We present Spatial-Temporal Interaction Networks (STIN) for compositional action recognition. Our model utilizes a generic detector and tracker to build objectgraph representations that explicitly include hand and constituent object nodes. We perform spatial-temporal reasoning among these bounding boxes to understand how the relations between subjects and objects change over time for a given action. By explicitly modeling the transformation of object geometric relations in a video, our model can effectively generalize to videos with unseen combinations of verbs and nouns.
Compositional action recognition over 174 classes. We present results when training the models using the detected bounding boxes.
Few-shot compositional action recognition on base categories and few-shot novel categories. We show results with object detection boxes.
Visualization of the detected bounding boxes
Paper
 |
|
Joanna Materzynska, Tete Xiao, Roei Herzig, Huijuan Xu*, Xiaolong Wang*, Trevor Darrell*
Something-Else: Compositional Action Recognition with Spatial-Temporal Interaction Networks
In CVPR, 2020.
(hosted on arXiv)
|
Prof. Darrell's group was supported in part by DoD, NSF, BAIR, and BDD. We would like to thank Fisher Yu and Haofeng Chen for helping set up the annotation pipeline. We would also like to thank Anna Rohrbach and Ronghang Hu for many helpful discussions, and Jonathon Luiten for comments on drafts. JM is supported by the Twenty Billion Neurons GmbH Fellowship.
