English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

The hypothesis that image datasets gathered online "in the wild" can producebiased object recognizers, e.g. preferring professional photography or certainviewing angles, is studied. A new "in the lab" data collection infrastructureis proposed consisting of a drone which captures images as it circles aroundobjects. Crucially, the control provided by this setup and the natural camerashake inherent to flight mitigate many biases. It's inexpensive and easilyreplicable nature may also potentially lead to a scalable data collectioneffort by the vision community. The procedure's usefulness is demonstrated bycreating a dataset of Objects Obtained With fLight (OOWL). Denoted as OOWL500,it contains 120,000 images of 500 objects and is the largest "in the lab" imagedataset available when both number of classes and objects per class areconsidered. Furthermore, it has enabled several of new insights on objectrecognition. First, a novel adversarial attack strategy is proposed, whereimage perturbations are defined in terms of semantic properties such as camerashake and pose. Indeed, experiments have shown that ImageNet has considerableamounts of pose and professional photography bias. Second, it is used to showthat the augmentation of in the wild datasets, such as ImageNet, with in thelab data, such as OOWL500, can significantly decrease these biases, leading toobject recognizers of improved generalization. Third, the dataset is used tostudy questions on "best procedures" for dataset collection. It is revealedthat data augmentation with synthetic images does not suffice to eliminate inthe wild datasets biases, and that camera shake and pose diversity play a moreimportant role in object recognition robustness than previously thought.

OOWL500: Overcoming Dataset Collection Bias in the Wild