The field of computer vision and image segmentation is moving towards leveraging foundation models, such as the Segment Anything Model (SAM), to improve performance in various applications, including medical imaging and object segmentation. Researchers are exploring innovative methods to adapt these models to specific domains and tasks, such as semi-supervised learning, domain adaptation, and few-shot learning. Notably, works like ConformalSAM and OP-SAM have demonstrated the potential of SAM-based approaches in semi-supervised semantic segmentation and one-shot polyp segmentation, respectively. Additionally, techniques like differentiable clustering and coalescent projections are being investigated to enhance the robustness and generalizability of these models. Overall, the field is witnessing significant advancements in developing more efficient, accurate, and adaptable image segmentation models.
Advances in Foundation Models for Image Segmentation
Sources
SuperCM: Improving Semi-Supervised Learning and Domain Adaptation through differentiable clustering
Depthwise-Dilated Convolutional Adapters for Medical Object Tracking and Segmentation Using the Segment Anything Model 2
FastSmoothSAM: A Fast Smooth Method For Segment Anything Model
Cross-Domain Few-Shot Learning with Coalescent Projections and Latent Space Reservation
ConformalSAM: Unlocking the Potential of Foundational Segmentation Models in Semi-Supervised Semantic Segmentation with Conformal Prediction
One Polyp Identifies All: One-Shot Polyp Segmentation with SAM via Cascaded Priors and Iterative Prompt Evolution
Families of Optimal Transport Kernels for Cell Complexes
CMP: A Composable Meta Prompt for SAM-Based Cross-Domain Few-Shot Segmentation
ScSAM: Debiasing Morphology and Distributional Variability in Subcellular Semantic Segmentation
Fully Automated SAM for Single-source Domain Generalization in Medical Image Segmentation
TextSAM-EUS: Text Prompt Learning for SAM to Accurately Segment Pancreatic Tumor in Endoscopic Ultrasound
MatSSL: Robust Self-Supervised Representation Learning for Metallographic Image Segmentation
Object segmentation in the wild with foundation models: application to vision assisted neuro-prostheses for upper limbs