Adaptive defect segmentation method improves metal 3d scan analysis
DCM-SAM: Defect-Conditioned Mixture of LoRA Experts for NPU-Deployed AM Defect Segmentation
Computer Vision and Pattern RecognitionMachine Learning
Summary
Metal parts made by 3D printing are checked for tiny defects using X-ray scans, but labeled data is scarce and defects are very small. The authors created DCM-SAM, a system that uses a frozen image model combined with small trained modules specialized for each defect type, trained only on synthetic images without needing user prompts. This method improves segmentation accuracy on real defect scans even though it was never trained on them. They also solved engineering challenges so the method runs efficiently on specialized hardware used in manufacturing machines.
What this means in practice
- •For industrial quality engineers: Deploy defect segmentation models on manufacturing machines to improve inspection accuracy for metal 3D printed parts without retraining on real images.
- •For embedded systems developers: Implement efficient segmentation models on NPUs for high-resolution defect detection in resource-constrained industrial hardware.
Authors
Md Mushfiqur Rahaman, Md Mahedi Hasan, Imtiaz Ahmed, Srinjoy Das
Abstract
Metal additive manufacturing parts are inspected by X-ray computed tomography, where labelled data is scarce, the pores and inclusions that matter span a few pixels, and inspection must happen at the machine. We present DCM-SAM, a defect-conditioned adaptive mixture of LoRA experts: one frozen Segment Anything backbone carries a separate Conv-LoRA expert bank and mask decoder per defect class, each trained in its own pass, without prompts, on synthetic slices alone, updating only 4.4% of the parameters. On benchmarks that XCT-SAM reports, DCM-SAM improves on every baseline for both classes from a ViT-B backbone against their ViT-H, and reaches 64.2% pore IoU on real NIST scans having seen no real images during training. Deployment then exposes what adaptation work rarely measures: on a Qualcomm Hexagon NPU, ViT-H and ViT-L compile yet cannot allocate at 1024x1024 image resolution, since activations rather than weights exceed the device ceiling, and quantizing weights does not help. ViT-B alone runs, but the adapted encoder then fails to allocate where the stock one succeeds, until a numerically identical rewrite of the attention lets the complete DCM-SAM run in FP16 at 1024x1024, with no operator falling back to the CPU, masks within 0.01% of pixels of the FP32 reference. Code: https://github.com/MushfiqShovon/DCM-SAM.