Problem
Research questions and friction points this paper is trying to address.
cervical cytology
Pap smear
cell detection
dense cell distribution
complex morphology
Center-Aware Detection with Swin-based Co-DETR Framework for Cervical Cytology
🤖 AI Summary
This work addresses the challenges of object detection in cervical cytology smears, where dense cell arrangements, complex morphologies, and fixed-size bounding box annotations hinder precise localization. To overcome these issues, the authors reformulate detection as a center-point prediction task and introduce a center-aware detection paradigm built upon the Co-DINO framework with a Swin-Large backbone. Their approach integrates center-preserving data augmentation, analytical geometric bounding box refinement, and track-specific loss function optimization. This methodology effectively mitigates localization jitter and achieves state-of-the-art performance, securing first place in Track B and second place in Track A of the RIVA Cervical Cytology Challenge. The results demonstrate significant improvements in detection accuracy and validate the efficacy and advancement of the proposed method for cytological image analysis.
📝 Abstract
Automated analysis of Pap smear images is critical for cervical cancer screening but remains challenging due to dense cell distribution and complex morphology. In this paper, we present our winning solution for the RIVA Cervical Cytology Challenge, achieving 1st place in Track B and 2nd place in Track A. Our approach leverages a powerful baseline, integrating the Co-DINO framework with a Swin-Large backbone for robust multi-scale feature extraction. To address the dataset's unique fixed-size bounding box annotations, we formulate the detection task as a center-point prediction problem. Tailoring our approach to this formulation, we introduce a center-preserving data augmentation strategy and an analytical geometric box optimization to effectively absorb localization jitter. Finally, we apply track-specific loss tuning to adapt the loss weights for each task. Experiments demonstrate that our targeted optimizations improve detection performance, providing an effective pipeline for cytology image analysis. Our code is available at https://github.com/YanKong0408/Center-DETR.
Innovation
Methods, ideas, or system contributions that make the work stand out.
center-aware detection
Swin-based Co-DETR
center-point prediction
geometric box optimization
data augmentation
🔎 Similar Papers
No similar papers found.
