SputumTB-ID: A Microscopic Sputum Image Dataset for Tuberculosis Detection

We present an image dataset of microscopic sputum images for Tuberculosis bacteria detection. The dataset was collected from 208 number of patients and captured 1029 images using 10x100 magnification using the built-in camera attached to the light microscope. Images obtained from the microscopy process were then captured in JPG format at a uniform size of 1280 x 960 pixels. A manual annotation was done by an expert from Rumah Sakit Imelda Pekerja Indonesia. The purpose of this dataset is support deep learning-based detection of small tuberculosis bacteria from microscopic sputum images. This dataset can be used directly for training YOLO.