LMS at ICCV 2023

Symbolic picture for the article. The link opens the image in a large view.
Source: Youssef Dawoud

This year our machine learning team participated in the International Conference on Computer Vision (ICCV) 2023 with two conference papers and one workshop paper. Our first paper is on “Out-of-Distribution Detection for Monocular Depth Estimation”. The paper presents an approach for estimating epistemic uncertainty at the pixel level. Second, we presented an approach on “Residual Pattern Learning for Pixel-Wise Out-of-Distribution Detection in Semantic Segmentation”. Here, an approach is presented that detects out-of-distribution pixels without compromising model performance. Third, our workshop paper “SelectNAdapt: Support Set Selection for Few-Shot Domain Adaptation”. Our work presents an approach to replace random support set selection with a more efficient mechanism that selects representative samples from the target domain of interest.

More information about each publication is available below. Note that the models and source code of all papers are publicly available. Many thanks to Youssef Dawoud (pictured) for taking the time to brief the group on the results of the conference.

Link to Out-of-Distribution Detection for Monocular Depth Estimation: https://openaccess.thecvf.com/content/ICCV2023/html/Hornauer_Out-of-Distribution_Detection_for_Monocular_Depth_Estimation_ICCV_2023_paper.html

Link to Residual Pattern Learning for Pixel-Wise Out-of-Distribution Detection in Semantic Segmentation: https://openaccess.thecvf.com/content/ICCV2023/html/Liu_Residual_Pattern_Learning_for_Pixel-Wise_Out-of-Distribution_Detection_in_Semantic_Segmentation_ICCV_2023_paper.html

Link to SelectNAdapt: Support Set Selection for Few-Shot Domain Adaptation: https://openaccess.thecvf.com/content/ICCV2023W/LIMIT/html/Dawoud_SelectNAdapt_Support_Set_Selection_for_Few-Shot_Domain_Adaptation_ICCVW_2023_paper.html