Fig. 7: Model prediction accuracy for FDM parts as a function of imaging sample size, ROI area, and pixel size.
From: Additive manufacturing source identification from photographs using deep learning
Points show model accuracy for document scanner images, indicating high accuracy for a variety of pixel and ROI sizes but degraded performance as sample size decreases. Example ROIs labeled in the plot are shown for the same FDM part; the top row shows four images with constant pixel size, and the bottom row shows four images with constant sample size. Contour lines show the imaging capability of various image sensors, indicating the achievable pixel size at a specific sample size based on the device specifications. Lens-based imaging methods have contour lines with a trade-off between pixel size and sample size, while scanners have horizontal lines at the highest possible resolution without any trade-off with sample size. Labels A–G on the plot refer to example images showing ROI size and resolution.
