CuVLER: Enhanced Unsupervised Object Discoveries through Exhaustive Self-Supervised Transformers
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866916157044293632 |
|---|---|
| author | Arica, Shahaf Rubin, Or Gershov, Sapir Laufer, Shlomi |
| author_facet | Arica, Shahaf Rubin, Or Gershov, Sapir Laufer, Shlomi |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2403_07700 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | CuVLER: Enhanced Unsupervised Object Discoveries through Exhaustive Self-Supervised Transformers Arica, Shahaf Rubin, Or Gershov, Sapir Laufer, Shlomi Computer Vision and Pattern Recognition 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. |
| title | CuVLER: Enhanced Unsupervised Object Discoveries through Exhaustive Self-Supervised Transformers |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2403.07700 |