Scrap Steel Image Dataset (China)

The dataset consists of 3440 labeled images with 29 label categories and a total of 6081 samples in VOC data format. Categories include various scrap steel types such as round steel, plates, H-steel, U-steel, engine blocks, brake discs, and more, organized by thickness ranges (2-5mm, 6-9mm, 10-12mm, 14-18mm, 20mm).

To download the dataset outside China, you can use Baidu Erranium service at https://baidu.erranium.com/. The dataset is approximately 9.75 GB and costs around $7 USD. The service allows downloading from Baidu Cloud without requiring a Chinese phone number or account registration.

Abstract

A Scrap Steel Image Dataset for Object Detection. The dataset consists of 3440 labeled images, 29 label categories and a total of 6081 samples in VOC data format. The dataset includes RGB-D image data with point clouds, depth images, and color images for each scene. In order to download the dataset outside China, you can use Baidu Erranium (paid service, approximately 9.75 GB for around $7 USD). The dataset is associated with the paper "An RGB-D-Based Thickness Feature Descriptor and Its Application on Scrap Steel Grading" published in IEEE Transactions on Instrumentation and Measurement (2023).