Assembly Line Parts Defect Visual Inspection Dataset

The current industrial manufacturing sector, especially assembly lines, faces challenges in efficiency and precision improvement. Traditional inspection methods rely on manual work, leading to inefficiency and prone to missed and false detection issues. Existing solutions often use simple machine vision technologies, which have drawbacks in detection range and adaptability. This dataset aims to enhance the defect recognition ability of machine vision in complex production environments to meet the high-precision quality inspection demands of smart manufacturing.Data acquisition uses high-resolution industrial cameras to capture high-resolution images of parts on a real production line, shot under various lighting and angles to ensure data diversity and authenticity. Data annotation goes through multiple rounds of labeling and expert review to ensure precision and consistency, with a ten-member team with mechanical and image processing backgrounds handling it. Preprocessing steps include image enhancement and noise removal, with specific algorithms used for standardization. The data is stored in JPG format and combined with database indexing for efficient search and access.The core advantage of the dataset is its high-quality annotation accuracy and consistency, with annotation errors below 3%. Innovative image augmentation techniques and automated quality assessment systems have been adopted, significantly enhancing the robustness and generalization of detection models. Compared to similar datasets, it has broader application adaptability, supporting various production line scenarios and filling the market gap of high-quality defect data resources. Additionally, the dataset is well-structured, allowing for flexible expansion and use in different detection tasks.