Conference item icon

Conference item : Poster

Diversified dynamic routing for vision tasks

Abstract:
Deep learning models for vision tasks are trained on large datasets under the assumption that there exists a universal representation that can be used to make predictions for all samples. Whereas high complexity models are proven to be capable of learning such representations, a mixture of experts trained on specific subsets of the data can infer the labels more efficiently. However using mixture of experts poses two new problems, namely (i) assigning the correct expert at inference time when a new unseen sample is presented. (ii) Finding the optimal partitioning of the training data, such that the experts rely the least on common features. In Dynamic Routing (DR) [21] a novel architecture is proposed where each layer is composed of a set of experts, however without addressing the two challenges we demonstrate that the model reverts to using the same subset of experts. In our method, Diversified Dynamic Routing (DivDR) the model is explicitly trained to solve the challenge of finding relevant partitioning of the data and assigning the correct experts in an unsupervised approach. We conduct several experiments on semantic segmentation on Cityscapes and object detection and instance segmentation on MS-COCO showing improved performance over several baselines.
Publication status:
Published
Peer review status:
Peer reviewed

Actions

Access Document

Files:
Publisher copy:
10.1007/978-3-031-25069-9_48

Authors

More by this author
Institution:
University of Oxford
Division:
MPLS
Department:
Engineering Science
Role:
Author


Publisher:
Springer
Host title:
Computer Vision – ECCV 2022 Workshops
Pages:
756-772
Series:
Lecture Notes in Computer Science
Series number:
13804
Place of publication:
Cham, Switzerland
Publication date:
2023-02-14
Acceptance date:
2022-06-19
Event title:
3rd Visual Inductive Priors for Data-Efficient Deep Learning Workshop (VIPriors 2022)
Event location:
Tel Aviv, Israel
Event website:
https://vipriors.github.io/2022/
Event start date:
2022-10-24
Event end date:
2022-10-24
DOI:
EISSN:
1611-3349
ISSN:
0302-9743
EISBN:
9783031250699
ISBN:
9783031250682


Language:
English
Keywords:
Subtype:
Poster
Pubs id:
1304014
Local pid:
pubs:1304014
Deposit date:
2022-11-14
ARK identifier:

Terms of use


Views and Downloads






If you are the owner of this record, you can report an update to it here: Report update to this record

TO TOP