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 Plus annual

Presents the access of premium content as premium feature

Premium Content

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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Recognizing artworks in a cultural site using images acquired from the user'spoint of view (First Person Vision) allows to build interesting applicationsfor both the visitors and the site managers. However, current object detectionalgorithms working in fully supervised settings need to be trained with largequantities of labeled data, whose collection requires a lot of times and highcosts in order to achieve good performance. Using synthetic data generated fromthe 3D model of the cultural site to train the algorithms can reduce thesecosts. On the other hand, when these models are tested with real images, asignificant drop in performance is observed due to the differences between realand synthetic images. In this study we consider the problem of UnsupervisedDomain Adaptation for object detection in cultural sites. To address thisproblem, we created a new dataset containing both synthetic and real images of16 different artworks. We hence investigated different domain adaptationtechniques based on one-stage and two-stage object detector, image-to-imagetranslation and feature alignment. Based on the observation that single-stagedetectors are more robust to the domain shift in the considered settings, weproposed a new method which builds on RetinaNet and feature alignment that wecalled DA-RetinaNet. The proposed approach achieves better results thancompared methods on the proposed dataset and on Cityscapes. To support researchin this field we release the dataset at the following linkhttps://iplab.dmi.unict.it/EGO-CH-OBJ-UDA/ and the code of the proposedarchitecture at https://github.com/fpv-iplab/DA-RetinaNet.

An Unsupervised Domain Adaptation Scheme for Single-Stage Artwork Recognition in Cultural Sites