MVDI25K dataset

*Introduction

The MVDI25K dataset is a large-scale dataset of microscopic vaginal discharge images, which contains 25,708 microscope images covering 10 object classes. The images in the dataset are collected from the HJ500 Discharge Analysis Workstation. Specifically, MVDI25K consists of RGB stored in 'jpg' formats, grayscale image annotations used for object segmentation, and bounding box annotations used for object detection including 'txt' and 'xml' formats.

The folder storage directory is as follows:

1. Dataset for object detection

MVDI25K
|——  JPEGImages
|        |——  *.jpg   （25708 images）
|        |——  *.txt   （25708 annotations in .txt）
|——  Annotations
|        |——  *.xml   （25708 annotations in .xml）

2.Dataset for object segmentation.

datasetforObjectSegmentaion
|——  Imgs
|        |——  *.jpg  （912 images）  
|——  mask
|        |——  *.png  （912 image annotation）

The 10 classes and corresponding category number in our dataset are as follows:
Leukocyte               0
Epithelial Cell         1
Candida1                2
Candida2                3 
Candida3                4 
Hyphae                  5
Trichomonas             6
Lactobacillus           7
Red Blood Cell          8
Clue Cell               9

Annotation format of 'txt' :
In the dataset, each object is annotated by a bounding box (BB), which can be denoted as (category, x1, y1, x2, y2). The "category" represents the category number, and the other four numbers represent the position of the object cells, that is, the coordinates of the upper-left and lower-right vertices of the object bounding box. Noted that the position coordinates here have been normalized.

*Usage License

All images and their associated annotations in MVDI25K can be used for academic purposes only, but any commercial use is prohibited.