"IC-Particle Dataset"

"The IC-Particle Dataset is a large-scale industrial microscopic image dataset designed for conductive particle detection, localization, and size measurement in advanced electronic packaging inspection scenarios. The dataset contains 5,640 high-resolution microscopic images collected from real industrial production environments, covering multiple challenging conditions, including normal distribution, dense adhesion, and defocus blur. Unlike conventional object detection datasets, IC-Particle focuses on quasi-circular micro targets with extremely small sizes, poor texture information, and severe particle overlap, which pose significant challenges for accurate detection and measurement. Each image is annotated with particle location information, while the dataset also supports weakly supervised learning based on center-point annotations. The dataset is specifically designed for tasks such as small object detection, weakly supervised localization, circle-aware detection, and unsupervised radius estimation. Owing to its diverse industrial conditions and high-density particle distributions, IC-Particle provides a valuable benchmark for evaluating the robustness, localization accuracy, and geometric measurement capability of modern deep learning models in industrial vision applications. The dataset can further promote research on precision inspection, intelligent manufacturing, and microscopic object analysis in advanced semiconductor and display packaging industries."