Open Stamped Parts Dataset

TL;DR

The Open Stamped Parts Dataset (OSPD) provides a large-scale, publicly available collection of real and synthetic stamped-hole images for auto manufacturing, enabling defect-detection research. Real data were captured with 7 cameras and annotated for holes, while synthetic data emulate the production environment with separate train/val/test splits and include bounding boxes and masks; 10% of holes are masked in synthetic data to simulate defects. A YOLOv7-based detector trained on synthetic data demonstrates promising performance with a modified recall of 67.2% and precision of 94.4%, illustrating the viability of synthetic data for manufacturing defect detection, while unpaired image-to-image translation offered limited improvements in this setup. The dataset and baseline methods aim to accelerate robust, real-time defect inspection on the factory floor and support transfer-learning research across synthetic-real domains.

Abstract

We present the Open Stamped Parts Dataset (OSPD), featuring synthetic and real images of stamped metal sheets for auto manufacturing. The real part images, captured from 7 cameras, consist of 7,980 unlabeled images and 1,680 labeled images. In addition, we have compiled a defect dataset by overlaying synthetically generated masks on 10\% of the holes. The synthetic dataset replicates the real manufacturing environment in terms of lighting and part placement relative to the cameras. The synthetic data includes 7,980 training images, 1,680 validation images and 1,680 test images, each with bounding box and segmentation mask annotations around all holes. 10\% of the holes in the synthetic data mimic defects generated in the real image dataset. We trained a hole-detection model on the synthetic-OSPD, achieving a modified recall score of 67.2\% and a precision of 94.4\% . We anticipate researchers in auto manufacturing use OSPD to advance the state of the art in defect detection of stamped holes in the metal-sheet stamping process. The dataset is available for download at: https://tinyurl.com/hm6xatd7.

Figures (7)

• Figure 1: Example of synthetically rendered (a.) and real stamped sheet metal (b.)
• Figure 2: a. Entrance side fixture with subset of cameras labeled. Exit side mirrors entrance side, but with even-numbered cameras b. Illustration of conveyor passing through fixture.
• Figure 3: Examples of masked holes, hole category DY1 (a.) and hole category DY2 (b.)
• Figure 4: Examples of labeled hole images, small holes (a.) and partially visible, low contrast holes (b.) Arrows point to the referenced hole annotations.
• Figure 5: Number of graded annotations by category for synthetic and real data.