Notes on a paper titled "Rethinking Computer-Aided Tuberculosis Diagnosis" presented at CVPR 2020.

在cvpr上少见的使用medical data的paper

Contributions

收集了一个规模宏大的TB X-ray数据集（TBX11K）。该数据集包含以下内容：

• 基于图像级别的标注以及基于边界框的TB区域标注。
• 基于图像级别的标注分为四类：健康、活动性结核、隐性结核以及非结核性病变

Rebuilding existing object detectors to execute simultaneous image classification and TB area detection (SSD, RetinaNet, Faster-RCNN, FCOS), simultaneously formulating evaluation metrics for both classification and detection. This serves as a baseline for the dataset.

Methods

The classification branch is capable of categorizin