CuVLER: Enhanced Unsupervised Object Discoveries through Exhaustive Self-Supervised Transformers
This work addresses unsupervised image segmentation, which is important for computer vision applications, but it appears incremental as it builds on existing self-supervised and graph-based techniques.
The paper tackles unsupervised object discovery by introducing VoteCut, which uses self-supervised models and graph partitioning for segmentation, and CuVLER, a zero-shot model trained with pseudo-labels, achieving significant improvements over previous state-of-the-art methods across multiple datasets.
In this paper, we introduce VoteCut, an innovative method for unsupervised object discovery that leverages feature representations from multiple self-supervised models. VoteCut employs normalized-cut based graph partitioning, clustering and a pixel voting approach. Additionally, We present CuVLER (Cut-Vote-and-LEaRn), a zero-shot model, trained using pseudo-labels, generated by VoteCut, and a novel soft target loss to refine segmentation accuracy. Through rigorous evaluations across multiple datasets and several unsupervised setups, our methods demonstrate significant improvements in comparison to previous state-of-the-art models. Our ablation studies further highlight the contributions of each component, revealing the robustness and efficacy of our approach. Collectively, VoteCut and CuVLER pave the way for future advancements in image segmentation.