Sacked Carton Dataset(SCD)
收藏资源简介:
Sacked Carton Dataset(SCD)是由华中科技大学数字制造装备与技术国家重点实验室创建的大型数据集,专注于物流行业的纸箱检测任务。该数据集包含来自三个不同场景(综合医药物流公司、电子商务物流公司和水果市场)的纸箱图像,共计520张。数据集的创建旨在解决由于不同场景下纸箱的标识、颜色、纹理等不一致导致的域偏移问题,从而提高模型在不同场景下的泛化能力。SCD数据集通过创新的图像合成方法,实现了源数据集中前景纹理与目标数据集纹理的替换,保持了前景对象与背景的上下文关系不变,有效增强了目标数据集。
The Sacked Carton Dataset (SCD) is a large-scale dataset developed by the State Key Laboratory of Digital Manufacturing Equipment and Technology, Huazhong University of Science and Technology, focusing on carton detection tasks in the logistics industry. It contains a total of 520 carton images collected from three distinct scenarios: comprehensive pharmaceutical logistics companies, e-commerce logistics companies, and fruit markets. This dataset was developed to address the domain shift problem caused by inconsistencies in carton markings, colors, textures and other attributes across different scenarios, so as to improve the generalization ability of models across various scenarios. The SCD dataset adopts an innovative image synthesis method to replace the foreground textures in the source dataset with those of the target dataset, while maintaining the contextual relationship between foreground objects and the background, effectively augmenting the target dataset.
数据集概述
数据集名称
CDS-FTR
数据集描述
Carton dataset synthesis method for loading-and-unloading carton detection based on deep learning
作者信息
- Gou, Lijun
- Wu, Shengkai
- Yang, Jinrong
- Yu, Hangcheng
- Lin, Chenxi
- Li, Xiaoping
- Deng, Chao
发表期刊
The International Journal of Advanced Manufacturing Technology
发表年份
2022
数据集使用环境
- Python 3.5
- OpenCV
- PIL (Pillow)
- Poisson Blending
- PyBlur
数据集运行脚本
-
测试表面分割: bash python test_get_surface.py -bf ./data/back -sf ./data/surface
-
测试轮廓构建: bash python test_construction_surface.py -bf ./data/back -sf ./data/construction
-
总结果: bash python main.py -ff ./data/fore/img -bf ./data/back -sf ./data/save




