Simplifying open-set video domain adaptation with contrastive learning

Giacomo Zara, Victor Guilherme Turrisi da Costa, Subhankar Roy, Paolo Rota, Elisa Ricci

Abstract: Highlights•The COLOSEO framework addressing open-set video domain adaptation with a contrastive learning approach.•A video-oriented temporal loss based on applying contrastive learning in order to model temporal consistency.•A thorough evaluation of the proposed framework on two benchmarks for open-set video domain adaptation.•Achievement of competitive or state-of-the-art results on all considered benchmarks.