裂缝缺陷专利成果运营数据库标注数据

本数据是评估若干专利进行专利运营时的内部测试数据脱敏后的目标检测数据集。数据包含多个生产和质检环节的图像，包括金属、复合材料等多种材料的表面。数据集包含了不同光照条件、表面纹理和颜色下的图像，可用于工业自动化检测、质量控制相关的目标检测模型训练。利用训练好的目标检测模型对生产线上的产品进行实时监控，快速识别并标记出金属、复合材料等表面的缺陷，从而及时采取措施，减少次品率，提高生产效率和产品质量。可以实现多种材料表面常见缺陷的高效识别，对工业自动化检测和质量控制具有重要意义。本数据是评估若干专利进行专利运营时的内部测试数据，图片工业生产线和质检环节的图像切片，包含了金属、复合材料等多种材料表面的图像。对图像预处理，保留原始分辨率，以保留材料表面的原有特征。采用随机旋转-随机缩放-随机亮度调整-随机对比度调整-随机噪声添加对图像进行增强。为提高模型对细微缺陷的识别能力，引入局部对比度增强算法：首先对原始图像应用高斯模糊，然后将原图像与模糊后的图像做差，得到局部细节图。最后，将这个局部细节图与原图像进行融合，融合程度由增强系数（本数据集中为0.5）控制，有效突出局部细节，有助于模型识别微小的裂缝和划痕。标注过程中使用LabelImg，每张图片只由一种缺陷，只打一个标签。数据标注采用CSV格式，filename为图像名，crack为裂缝，dent为凹痕，missing-head为缺头，paint-off为脱漆，scratch为划痕。缺陷存在标记为1，不存在标记为0。标注时只关注主要缺陷，忽略次要或不明显的问题。如果图里有多个缺陷，按最严重的标。

This dataset is a de-identified object detection dataset derived from internal test data used for patent operation evaluation of several patents. It contains images collected from multiple production and quality inspection links, covering surfaces of various materials including metals, composite materials and other materials. The dataset includes images captured under diverse lighting conditions, surface textures and colors, and can be utilized for training object detection models for industrial automated inspection and quality control applications. Trained object detection models can perform real-time monitoring of products on production lines, rapidly identify and mark defects on surfaces of metals, composite materials and other materials, enabling timely corrective actions, reducing defective product rates, and improving production efficiency and product quality. This dataset enables efficient recognition of common defects on surfaces of various materials, holding great significance for industrial automated inspection and quality control. This dataset is internal test data for patent operation evaluation of several patents, consisting of image slices from industrial production lines and quality inspection links, including images of surfaces of various materials such as metals and composite materials. Image preprocessing was conducted while retaining the original resolution to preserve the original features of the material surfaces. Image augmentation was implemented using a combination of random rotation, random scaling, random brightness adjustment, random contrast adjustment and random noise addition. To improve the model's capability to recognize subtle defects, a local contrast enhancement algorithm was adopted: first, Gaussian blur was applied to the original image, then the difference between the original image and the blurred image was calculated to obtain a local detail map. Finally, this local detail map was fused with the original image, with the fusion degree controlled by the enhancement coefficient (set to 0.5 in this dataset), which effectively highlights local details and aids the model in detecting tiny cracks and scratches. During the annotation process, LabelImg was used. Each image corresponds to only one type of defect, and only one annotation tag is applied per image. The annotations are stored in CSV format, where the "filename" field represents the image name, and the defect tags are crack, dent, missing-head, paint-off and scratch. A defect is marked as 1 if it exists, and 0 otherwise. During annotation, only primary defects are focused on, while secondary or inconspicuous issues are disregarded. If multiple defects appear in an image, the most severe one will be annotated.

搜集汇总

数据集介绍

特点

该数据集包含606条工业生产线和质检环节的图像切片，用于工业自动化检测和质量控制。数据标注采用CSV格式，标注了裂缝、凹痕、缺头、脱漆和划痕等缺陷，标注规则严格，每张图片只标注一种主要缺陷。