Datasets
Standard Dataset
Instance Segmentation for Droplet Through Sessile Drop Technique
- Citation Author(s):
- Submitted by:
- Devanshu Sawant
- Last updated:
- Wed, 12/18/2024 - 16:24
- DOI:
- 10.21227/11fh-6w04
- Data Format:
- License:
42 Views- Categories:
- Keywords:
Abstract
This paper describes a dataset of droplet images captured using the sessile drop technique, intended for applications in wettability analysis, surface characterization, and machine learning model training. The dataset comprises both original and synthetically augmented images to enhance its diversity and robustness for training machine learning models. The original, non-augmented portion of the dataset consists of 420 images of sessile droplets. To increase the dataset size and variability, an augmentation process was applied, generating 1008 additional images. This augmentation employed adjustments to brightness (±18%) and exposure (±10%), simulating variations in lighting conditions during image acquisition. The combined dataset, totaling 1428 images, offers a valuable resource for developing and evaluating automated droplet analysis methods, particularly those based on deep learning. This abstract provides an overview of the dataset's composition and augmentation strategy, highlighting its potential contribution to the field of surface science and image analysis.
These instructions are for using the sessile droplet image dataset, which is formatted for use with YOLO (You Only Look Once) object detection models. The dataset contains both original and augmented images of sessile droplets.
Dataset Structure:
The dataset is organized as follows:
- Images: Images are stored in a directory (e.g.,
images). - Labels: Corresponding bounding box annotations are stored in a separate directory (e.g.,
labels). - Data Split: The dataset is split into training, validation, and test sets. This split is defined in a configuration file (e.g.,
data.yaml).
File Formats:
- Images: Images are in
.jpgor.pngformat. - Labels: Labels are in
.txtfiles, one per image. The filename of the label file matches the corresponding image filename (e.g.,image001.jpghas a corresponding label fileimage001.txt). - Data Configuration (data.yaml): This YAML file specifies the paths to the training, validation, and test sets, as well as the number of classes and class names.
Label File Format (YOLO):
Each line in a label .txt file represents a single object (droplet) and follows the YOLO format:
<class_id> <x_center> <y_center> <width> <height>
Where:
<class_id>: Integer representing the object class. In this dataset, there is only one class (droplet), so<class_id>is always0.<x_center>: Normalized x-coordinate of the center of the bounding box.<y_center>: Normalized y-coordinate of the center of the bounding box.<width>: Normalized width of the bounding box.<height>: Normalized height of the bounding box.
Normalization is performed by dividing the pixel coordinates by the image width and height, respectively. The values are between 0 and 1.
