cs.CVAug 12, 2026

Low Cost Two-Stage Fabric Defect Detection at the Edge

Authors: Rasel Hossen, Diptajoy Mistry, Mosaddek Hossain Kamal

Organizations: Department of Computer Science and Engineering, University of Dhaka, Dhaka, Bangladesh

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Fabric inspection in the garment industries of low-income economies remains largely manual, and commercial vision systems are priced beyond most small and medium mills. Because defects are sparse under controlled production, a natural response is a cascade: screen every frame with a cheap anomaly detector and invoke a full detector only on suspicious frames. We build such a cascade for four knit-fabric defect classes and deploy it end-to-end on an NVIDIA Jetson Nano with TensorRT FP16. Stage 1 is a compact convolutional autoencoder with decoder attention gates, an edge-weighted reconstruction loss, and feature-level distillation from a frozen YOLOv5n teacher; Stage 2 is YOLOv5n, invoked only on flagged frames. On a 249-image benchmark disjoint from detector training (20 defective, 229 non-defective), Stage 1 at a recall-prioritised threshold flags all 20 defective images (95% CI 0.83-1.00) at a false-positive rate of 49.3% (113/229), reducing false positives by 19.3% relative to a plain autoencoder (p=0.011). The parallel pipeline reaches 13.45 FPS against 9.86 FPS for a sequential YOLO-only loop. Our central finding comes from decomposing that 1.36x: 91% of it is attributable to overlapping JPEG decode with inference rather than to the cascade, which contributes only a 5.1% inference reduction at the measured forwarding rate p = 0.534. We further show that forwarding here is false-positive-limited rather than prevalence-limited - 85% of forwarded frames are false alarms - and quantify the 29-45% inference reduction attainable under tighter calibration. We report this as a caution for cascade speedups measured without controlling the data path, and position the system as AI-assisted triage rather than autonomous acceptance.

Figures & tables

Figure 1: Two-stage fabric defect detection (logical flow). Stage 1 filters normal frames; Stage 2 runs only on flagged anomalies. Deployment uses a pipelined producer–consumer layout (Section 4.4 ), not strict AE-then-YOLO serialisation on every frame.
Figure 2: System architecture with per-stage latency. Camera input is preprocessed and screened by Stage 1 (AE). Only anomalous frames are forwarded to Stage 2 (YOLO) for localisation and classification.
Figure 3: The four defect types with in-image labels.
Figure 4: Base autoencoder. Encoder compresses input xx to latent yy ; decoder reconstructs x~\tilde{x} ; training minimises L=∥x−x~∥2\mathcal{L}=\|x-\tilde{x}\|_{2} .
Figure 5: Modified autoencoder: encoder, bottleneck, and decoder with attention gates.
Figure 6: Distillation framework. A frozen YOLOv5n teacher provides defect-aware features, projected to match the autoencoder’s latent space. The autoencoder aligns its latent representation with these features while minimising reconstruction error. At inference, only the autoencoder (left branch) runs.
SourceImagesNotes
Fabric Detection [ 39 ]89Roboflow (DENEME)
ISL-Knit [ 38 ]2,587Bangladesh knit dyeing
Fabric [ 40 ]76Roboflow (Afrasiab)
defect-needle [ 41 ]32Roboflow (ecovers)
Fabric Defect Augmented [ 42 ]209Roboflow (TrainerWork)
Multifab (real)195Bangladesh garment factory
Table 1: Dataset composition by source (4,188 images). Real and synthetic in-house images are listed separately; no synthetic image is used for evaluation.
ClassInstancesShare
Fly yarn1,6818.4%
Hole2,82514.0%
Oil spot13,22465.7%
Thread error2,39911.9%
Total20,129100%
Table 2: Defect class distribution (instance counts; 20,129 total)
ThresholdRecall (defects)Overall accuracy
0.0009 (low)80% (16/20)10.4%
0.12 (high)35% (7/20)55.0%
Table 3: Canny Stage 1 on the 249-image benchmark. Neither threshold is usable: the low setting flags nearly all normal frames, the high setting misses most defects.
ModelRecallFPFPRAcc.
Base AE19/20140/2290.6110.434
 [0.75, 1.00] [0.54, 0.67] 
Modified AE20/20113/2290.4930.546
 [0.83, 1.00] [0.43, 0.56] 
Always-normal0/200/2290.0000.920
Table 4: Base vs. modified autoencoder on the 249-image benchmark (20 defective, 229 non-defective) at τ=0.00245\tau=0.00245 . FPR = false-positive rate on the 229 non-defective images. Brackets are Clopper–Pearson 95% intervals. The final row gives the trivial classifier that labels every frame normal, included because the accuracy column is otherwise easy to misread.
Figure 7: Base vs. modified autoencoder: defects found (of 20), false positives, and normal images correctly passed (of 229). The false-positive reduction is the supported effect; the one-image recall difference is not.
ModelPar.GFLOPsmAP50mAP50-95Prec.Rec.Size
YOLOv5n1.76M4.10.7420.3630.8250.6373.7 MB
YOLOv5s7.02M15.80.7160.3450.7750.65314.3 MB
YOLOv8n3.01M8.10.6760.3390.7590.6046.2 MB
Table 5: Stage-2 candidates at 320×320320\times 320 . Box-level metrics on the validation split.
ModelTNFPFNTPAcc.95% CI
YOLOv5n2821171199/202 = 98.5%[0.957, 0.997]
YOLOv5s2910172201/202 = 99.5%[0.973, 1.000]
YOLOv8n30020152182/202 = 90.1%[0.851, 0.939]
Table 6: Image-level performance on the 202-image validation split (172 defective, 30 clean), with confusion-matrix counts and Clopper–Pearson 95% intervals on accuracy. YOLOv5n and YOLOv5s differ by two images; YOLOv8n is separated from both by its 20 false negatives.
Figure 8: YOLOv5n detection example ( 320×320320\times 320 ): bounding boxes and confidence scores for the four defect classes.
Figure 9: Model size versus [email protected] for the Stage-2 candidates. YOLOv5n is selected for footprint; the mAP differences among the three are within the resolution of the validation split.
ConfigurationInf.WallOverheadFPS
 (ms)(ms)(ms) 
YOLO-only (sequential)47.77101.4053.639.86
AE-only (sequential)27.2982.2554.9612.16
Two-stage (parallel)45.3674.3328.9713.45
Table 7: Jetson Nano runtime on the 249-image benchmark (TensorRT FP16, τ=0.00245\tau=0.00245 , p=0.534p=0.534 ). Inference for the two-stage row is tAE+p tYOLO=22.67+0.534(42.47)t_{\mathrm{AE}}+p\,t_{\mathrm{YOLO}}=22.67+0.534(42.47) . The overhead column is wall minus inference and is the quantity that turns out to matter.
Figure 10: Wall time decomposed. The inference bars are nearly equal (47.77 vs. 45.36 ms): the cascade saves little detector compute at p=0.534p=0.534 . Almost the entire difference in total height is non-inference time, which the parallel runtime overlaps and the sequential baseline does not.
 FPRppInf. (ms)vs. YOLO
Measured ( τ=0.00245\tau{=}0.00245 )0.4930.53445.35−5.1%-5.1\%
Projected0.2000.26433.89−29.0%-29.0\%
Projected0.1000.17229.99−37.2%-37.2\%
Projected0.0500.12628.03−41.3%-41.3\%
Projected0.0100.09026.47−44.6%-44.6\%
Table 8: Projected forwarding rate and inference cost at π=0.0803\pi=0.0803 as a function of Stage-1 false-positive rate, from Eq. ( 6 ) with tAE=22.67t_{\mathrm{AE}}=22.67 ms and tYOLO=42.47t_{\mathrm{YOLO}}=42.47 ms. Only the first row is measured; the rest assume recall is preserved and are upper bounds on the achievable benefit.

Explore similar work

CardsList
Jun 3, 2026

Real-Time Industrial Defect Detection on Edge Hardware Using Fine-Tuned YOLOv8: A Systematic Benchmark on the NEU Surface Defect Database and MVTec AD with Automotive & Battery Manufacturing Extensions

Automated surface defect detection is critical for ensuring rigorous quality control in high-speed manufacturing environments. While deep learning models offer remarkable accuracy, deploying them on resource-constrained edge hardware without introducing significant latency remains a persistent challenge. This paper presents Industrial-YOLO, an edge-optimized framework built upon a fine-tuned YOLOv8 architecture specifically engineered for real-time industrial defect detection. We conduct a systematic benchmark utilizing the NEU surface defect database for steel sheets and the MVTec AD dataset, supplemented with custom automotive manufacturing extensions representing real-world structural anomalies (scratches, pits, and inclusions). To bridge the gap between algorithmic complexity and edge hardware constraints, target-specific optimizations are introduced via TensorRT and OpenVINO acceleration engines. Experimental results demonstrate that Industrial-YOLO achieves a high-velocity inference speed exceeding 120 FPS on the NVIDIA Jetson Orin platform while maintaining an exceptional mean Average Precision (mAP) of 98.5%. The proposed framework showcases highly robust, zero-latency performance when deployed directly onto an active automotive assembly line, offering a scalable blueprint for next-generation automated optical inspection (AOI) systems.

Emmanuel Ezeji Somtochukwu, Nitesh Rijal

Zema AI Labs, Abuja, Nigeria

Edge InferenceIndustrial InspectionObject DetectionReal-Time Object Detection
AskPDF
Mar 16, 2026

GATE-AD: Graph Attention Network Encoding For Few-Shot Industrial Visual Anomaly Detection

Few-Shot Industrial Visual Anomaly Detection (FS-IVAD) comprises a critical task in modern manufacturing settings, where automated product inspection systems need to identify rare defects using only a handful of normal/defect-free training samples. In this context, the current study introduces a novel reconstruction-based approach termed GATE-AD. In particular, the proposed framework relies on the employment of a masked, representation-aligned Graph Attention Network (GAT) encoding scheme to learn robust appearance patterns of normal samples. By leveraging dense, patch-level, visual feature tokens as graph nodes, the model employs stacked self-attentional layers to adaptively encode complex, irregular, non-Euclidean, local relations. The graph is enhanced with a representation alignment component grounded on a learnable, latent space, where high reconstruction residual areas (i.e., defects) are assessed using a Scaled Cosine Error (SCE) objective function. Extensive comparative evaluation on the MVTec AD, VisA, and MPDD industrial defect detection benchmarks demonstrates that GATE-AD achieves state-of-the-art performance across the 11- to 88-shot settings, combining the highest detection accuracy (increase up to 1.8%1.8\% in image AUROC in the 8-shot case in MPDD) with the lowest per-image inference latency (at least 25.05%25.05\% faster), compared to the best-performing literature methods. In order to facilitate reproducibility and further research, the source code of GATE-AD is available at https://github.com/gthpapadopoulos/GATE-AD.

Aggelos Psiris, Yannis Panagakis, Maria Vakalopoulou +1

Harokopio University of Athens, Athens, Greece · National and Kapodistrian University of Athens, Athens, Greece · Archimedes, Athena Research Center, Athens, Greece +1

Few-Shot Anomaly DetectionIndustrial Anomaly DetectionReconstruction-Based Anomaly DetectionGraph Attention Networks
AskPDF
Apr 22, 2026

Accelerating New Product Introduction for Visual Quality Inspection via Few-Shot Diffusion-Based Defect Synthesis

Industrial visual inspection systems often suffer from a severe scarcity of labeled defect data, particularly during the early stages of New Product Introduction (NPI). This limitation hinders the deployment of robust supervised detectors precisely when automated quality control is most needed. We present an end-to-end generative framework for high-fidelity, few-shot defect synthesis that enables both in-domain augmentation and cross-domain transfer. Our approach disentangles defect morphology from background appearance by combining masked textual inversion for defect representation learning, noise-blended conditioned generation for surface-aware synthesis, and gradient-aware post-processing for seamless visual integration. We evaluate the framework in two practically relevant settings: few-shot data augmentation, where synthetic samples enrich a small set of real defects, and zero-shot adaptation, where defects learned from a source domain are transferred to a novel target surface without any real target-domain defect examples. Using RF-DETR as the downstream detector, we show that the proposed pipeline substantially narrows the domain gap on a private industrial dataset. In the few-shot setting, synthetic augmentation improves mAP from 78.8% to 83.3%. In the zero-shot setting, synthetic domain adaptation improves mAP from 65.0% to 85.1%. These results demonstrate that high-fidelity defect synthesis can meaningfully accelerate NPI by enabling effective inspection models before sufficient real defect data has been collected.

Serkan Hamdi Güğül, Kemal Levi, Burak Acar

Relimetrics, Inc.

Synthetic-to-Real Domain AdaptationSynthetic Data AugmentationIndustrial InspectionDiffusion-Based Image GenerationSynthetic Anomaly Generation
AskPDF