This dataset contains 2,638 high-resolution powder bed (PB) images and 2674 optical tomography (OT) images with visible defects, collected from the EOS M290 machine during the laser-based powder bed fusion (PBF-LB) process. The PB images were collected by the EOSTATE PowderBed monitoring system before and after each layer's exposure. The OT images provide layer-wise thermal and structural insights during the build. All data were sourced from ADDLAB at Aalto University. All samples were printed with 316L stainless steel powder (Dv(25) = 22 µm, Dv(50) = 37 µm, Dv(75) = 58 µm), recoating speed 80 mm/s. Detailed process parameters for each set are: Set Laser Power (W) Scan Speed (mm/s) Hatch Distance (mm) Layer Thickness (µm) Volumetric Energy Density (J/mm³) 1 80 - 250 780 - 3120 0.09 40 8.89, 17.81, 26.71, 35.61 2 215 1083 0.09 40 55.15 3 195 1083 0.09 20 100.03 In addition to the raw images, the dataset includes manually labeled annotations and preprocessed and cropped images suitable for the training and validation of machine learning and deep learning models: Original Images and Manual Labels: Includes 2,638 raw powder bed (PB) images (JPEG, 1280×1024 pixels) and 2,674 optical tomography (OT) images (JPEG, 2000×2000 pixels), along with their corresponding manual annotations in XML format, created using LabelImg. Cropped Defects and Classification: Provides defect-centered image crops with binary class labels ("Good" vs "Defects") for image classification tasks. Preprocessed Data: Includes resized and normalized images, cropped to consistent dimensions and ready for data augmentation and model training. This is designed for defect detection using object detection models. This image dataset can be used for image recognition, object detection, and other potential research in Metal Additive Manufacturing.

2025