MARVEL’s partner FBK, participated in the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) that took place from 2 to 8 January 2023, in Waikoloa, Hawaii, USA.

In more detail, on January 7th, 2023, Luca Zanella presented a paper, also coauthored by Giulio Mattolin, Elisa Ricci, and Yiming Wang. The paper is titled “ConfMix: Unsupervised Domain Adaptation for Object Detection via Confidence-based Mixing” and contributes to exploring how sample-mixing can be applied to adaptive object detection.

As known, object detection is a fundamental task in computer vision, the first step in many applications from traffic monitoring to video surveillance. Detectors rely on deep learning, but they often suffer from severe performance degradation when being tested on images that are visually different from the ones encountered during training, due to the domain shift. This scientific contribution addresses the problem of devising an original solution for Unsupervised Domain Adaptation (UDA). It is inspired by sophisticated data augmentation techniques that synthesize mixed samples with target and source images to improve the generalisation ability of deep architectures. Not a trivial task!

If you want to know more feel free to read the paper here or explore the code here