焊点缺陷检测数据

焊点缺陷检测数据是我公司通过焊接机器人对金属表面焊接焊点在相对于相机的不同位置拍摄。在每个位置，每个焊接焊点都以5种不同的旋转方式捕获。这为每个焊接焊点创建了相对于相机的独特方向。每个独特的方向被捕获4次并编码用于训练，然后捕获另一个焊接焊点进行编码并进行比对测试。因此，在训练中使用的焊接焊点与测试中使用的焊接焊点都不相同。数据集中的每个焊接焊点都有训练图像和测试图像，从而创建了500/625的总体训练/测试图像数据。

This soldering joint defect detection dataset was collected by our company using welding robots, which captured soldered joints on metal surfaces at various positions relative to the camera. For each position, each soldered joint was captured with 5 distinct rotational orientations, creating a unique relative orientation between each joint and the camera. Each unique orientation was captured 4 times and encoded for model training, after which a new distinct soldered joint was captured, encoded and used for comparative testing. Therefore, the soldered joints used for training are entirely distinct from those utilized for testing. Each soldered joint in the dataset has both training and test images, resulting in an overall training/test image dataset with a split of 500/625.