The field of computer vision is rapidly advancing, with a focus on improving segmentation and object tracking. Recent developments have led to the creation of more efficient and effective algorithms for tasks such as multi-object tracking, document image segmentation, and instance segmentation. These advancements have the potential to be applied in a variety of real-world applications, including autonomous vehicles, document analysis, and medical imaging. Noteworthy papers include DocSAM, which presents a unified framework for document image segmentation, and SAM2MOT, which introduces a novel paradigm for multi-object tracking by segmentation. These papers demonstrate significant improvements in accuracy, efficiency, and adaptability, and highlight the potential for future advancements in the field.
Advances in Computer Vision and Segmentation
Sources
Page Classification for Print Imaging Pipeline
TQD-Track: Temporal Query Denoising for 3D Multi-Object Tracking
DocSAM: Unified Document Image Segmentation via Query Decomposition and Heterogeneous Mixed Learning
CoMBO: Conflict Mitigation via Branched Optimization for Class Incremental Segmentation
A Novel Cholesky Kernel based Support Vector Classifier
SAM2MOT: A Novel Paradigm of Multi-Object Tracking by Segmentation
CMaP-SAM: Contraction Mapping Prior for SAM-driven Few-shot Segmentation
BoxSeg: Quality-Aware and Peer-Assisted Learning for Box-supervised Instance Segmentation
S^4M: Boosting Semi-Supervised Instance Segmentation with SAM
Turin3D: Evaluating Adaptation Strategies under Label Scarcity in Urban LiDAR Segmentation with Semi-Supervised Techniques