Description

Deep metric learning learns a nonlinear embedding of the data by using deep neural networks. The embedding preserves the distance between similar data points close and dissimilar data points far on the embedding space. Contrastive learning is currently the most successful framework for deep metric learning. The underlying idea is to define losses contrasting positive pairs from sets of negative samples.
We will show the effectiveness of the proposed approach in the context of action segmentation of untrimmed videos, that is the process of partitioning a video into semantically and temporally homogeneous events.
Advances in automatic acton segmentation in untrimmed videos would benefit the understanding and contextualizing of a video

The work is under the scope of the following projects:

• GREAT: Beyond Graph Neural Networks: Joint graph topology learning and graph-based inference for computer vision (web)