arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2406.08046cs.CVeess.SP

使用 Swin Transformer 和 RT-DETR 的无线胶囊内镜出血帧分类与检测鲁棒流水线

A Robust Pipeline for Classification and Detection of Bleeding Frames in Wireless Capsule Endoscopy using Swin Transformer and RT-DETR

IIT Tirupati（蒂鲁帕蒂印度理工学院）

Sasidhar Alavala, Anil Kumar Vadde, Aparnamala Kancheti, Subrahmanyam Gorthi

阅读 PDF
arXiv 页面 ↗HTML 原文 ↗

AI总结：

面向 WCE 出血帧识别，提出融合 Lab/CLAHE/高斯模糊预处理、Swin Transformer 分类与 RT-DETR 检测的鲁棒流水线，验证集分类准确率达 98.5%、AP50 达 66.7%。

摘要：

本文介绍了我们参加 Auto WCEBleedGen Challenge V2 2024 的方法。我们的解决方案将用于出血帧初步分类的 Swin Transformer 与用于无线胶囊内镜（WCE）中出血进一步检测的 RT-DETR 相结合，并通过一系列图像预处理步骤加以增强。这些步骤包括将图像转换到 Lab 色彩空间、应用对比度受限自适应直方图均衡化（CLAHE）以改善对比度，以及使用高斯模糊抑制伪影。Swin Transformer 采用带有移位窗口的分层架构，以高效管理自注意力计算，既聚焦局部窗口，又支持跨窗口交互。RT-DETR 具有用于快速处理多尺度特征的高效混合编码器，以及用于提升精度的最小不确定性查询选择。由 Ablation-CAM 生成的类激活图与模型决策相符。在验证集上，该方法取得 98.5% 的分类准确率（在其他最先进模型中最高），而未进行任何预处理时为 91.7%；其 AP50 为 66.7%，而最先进的 YOLOv8 为 65.0%。在测试集上，该方法取得的分类准确率和 F1 分数分别为 87.0% 和 89.0%。

英文摘要：

In this paper, we present our approach to the Auto WCEBleedGen Challenge V2 2024. Our solution combines the Swin Transformer for the initial classification of bleeding frames and RT-DETR for further detection of bleeding in Wireless Capsule Endoscopy (WCE), enhanced by a series of image preprocessing steps. These steps include converting images to Lab colour space, applying Contrast Limited Adaptive Histogram Equalization (CLAHE) for better contrast, and using Gaussian blur to suppress artefacts. The Swin Transformer utilizes a tiered architecture with shifted windows to efficiently manage self-attention calculations, focusing on local windows while enabling cross-window interactions. RT-DETR features an efficient hybrid encoder for fast processing of multi-scale features and an uncertainty-minimal query selection for enhanced accuracy. The class activation maps by Ablation-CAM are plausible to the model's decisions. On the validation set, this approach achieves a classification accuracy of 98.5% (best among the other state-of-the-art models) compared to 91.7% without any pre-processing and an $\text{AP}_{50}$ of 66.7% compared to 65.0% with state-of-the-art YOLOv8. On the test set, this approach achieves a classification accuracy and F1 score of 87.0% and 89.0% respectively.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。

↑