Unsupervised Object Discovery: A Comprehensive Survey of Methods and Taxonomy
This survey provides a comprehensive overview of unsupervised object discovery methods, categorizing them based on tasks (clustering, localization, segmentation, decomposition) and techniques, highlighting challenges and future directions in this evolving field.
Unsupervised Object Discovery through Masked Multi-Query Slot Attention
Unsupervised object discovery can be improved by selectively masking background regions during training and using multi-query slot attention to learn more stable and generalizable object representations.
CuVLER: Enhanced Unsupervised Object Discoveries through Exhaustive Self-Supervised Transformers
VoteCut and CuVLER revolutionize unsupervised object discovery with self-supervised transformers.
CuVLER: Enhanced Unsupervised Object Discoveries through Exhaustive Self-Supervised Transformers
The authors introduce VoteCut and CuVLER, innovative methods for unsupervised object discovery and segmentation, showcasing significant improvements over previous state-of-the-art models.
© 2024 by Linnk AI