Development of a Customized Image Set for Object Detection in a Materials Synthesis Lab Using Convolutional Neural Networks

Abstract

Convolutional Neural Networks (CNNs) have shown remarkable performance in object detection tasks, especially when trained on large and diverse datasets. However, in specialized domains such as materials synthesis laboratories, generic datasets may not capture the specific objects of interest or the unique challenges of the environment. This paper presents the development of a custom image dataset tailored for object detection in a materials synthesis laboratory. The dataset includes annotated images of equipment, chemicals, and other objects commonly found in such environments. We also describe the process of collecting and labeling the dataset, including the challenges faced and the strategies used to address them. To demonstrate the utility of the dataset, we trained a CNN model using the popular YOLO (You Only Look Once) architecture and evaluated its performance on a test set. The results show that our custom dataset enables the CNN model to accurately detect objects in materials synthesis laboratories, highlighting the importance of domain-specific datasets for enhancing the performance of object detection systems in specialized environments.