White button mushroom RGB-D Images

Published: 02-11-2020| Version 1 | DOI: 10.17632/4bdkk5wwjw.1
Contributor:
jingjun tsao

Description

This dataset is consisted with 1150 pairs of RGB and depth white button mushroom image. The RGB and depth image pairs are aligned. We label the white button mushroom objects in the images to get the annotations file. In each annotation file, the content in the first line is the number of objects in the corresponding image. The content in other lines are the bounding box coordinates (x1,y1,x2,y2).

Files

Steps to reproduce

1.We use Intel RealSense D435 to collect experimental data. 2.The depth value of corresponding pixel is 16 bits. In order to save depth value accurately, we collect white button video as bag form firstly. 3.we extract RGB and depth frames of white button mushroom from collected video at certain time interval. We align the RGB and depth images during the extracting process. 4.Filtering out ineffective frames such as blur and incomplete depth information frame is also necessary. 5.Dividing the images into training , validation and test set. Then we can use the data to train model.