FDM Pore Segmentation Dataset for In-Situ 3D Printing Monitoring

This dataset contains annotated pore segmentation masks extracted from in-situ video recordings of extrusion-based fused deposition modeling (FDM) 3D printing. The data were collected during printing of chicken-feather-fiber reinforced polybutylene succinate (CFF/PBS) composite filaments. Video was captured using an endoscope camera mounted near the print head to monitor layer-by-layer extrusion and pore formation. Frames were extracted from a 17-minute video recorded at 1280×720 resolution and 30 FPS. After temporal sampling, 1,607 frames were selected and manually annotated using polygon masks to precisely label pores. The dataset is provided in multiple formats to support different machine learning frameworks: • YOLOv8 segmentation format• COCO segmentation format• Pascal VOC format• TensorFlow TFRecord format• SAM2 compatible format This dataset accompanies the paper: "Automated Pore Detection from In-Situ FDM 3D Printing Video: A Comparative Evaluation of Modern Segmentation Models" Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2026. If you use this dataset in your research, please cite the following paper: @InProceedings{Al_Ahad_Khan_2026_WACV, author = {Al Ahad Khan, Abdullah and Islam, Md Shariful and Li, Lin and Jiang, Lai and Ghaffari, Noushin}, title = {Automated Pore Detection from In-Situ FDM 3D Printing Video: A Comparative Evaluation of Modern Segmentation Models}, booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)}, month = {March}, year = {2026}, pages = {4673--4681} }