Datasets
Standard Dataset
Wafer Surface Defect
- Citation Author(s):
-
MengyunLi
China Jiliang University - Submitted by:
- Mengyun Li
- Last updated:
- Wed, 10/16/2024 - 22:38
- DOI:
- 10.21227/7x9f-zt52
- License:
111 Views- Categories:
- Keywords:
Abstract
The dataset is a self-constructed wafer surface defect dataset, with each image captured in real-time. The extraction and segmentation of wafer image have been performed, and each image represents a single individual die. The dataset primarily includes images of defect-free dies, as well as four types of defective images: particle, scratch, stain, and liquid residual. A total of 500 images are included, and the various types of defects within the images have been annotated using the Make Sense online annotation tool. The location and category information of the defects are stored in YOLO format as txt files. Each image has a resolution of 680×680 pixels.
In this dataset, the images and labels are stored in two separate folders named "images" and "labels", respectively. Each txt file in the "labels" folder contains the class of the target and the coordinates of the bounding box, with the format as follows:
[class x_center y_center w h]
The class denotes the category ID of the target. x_center and y_center represent the coordinates of the center of the bounding box, while w and h refer to the width and height of the bounding box, respectively.
