Efficient Self-Filtered PatchCore for Unsupervised Fabric Defect Detection
Memory-bank compression, cross-dataset validation & edge deployment
I work under Senior Lecturer Shahed Alam on unsupervised fabric-defect detection with PatchCore and deep visual features, using reproducible PyTorch training, evaluation and ablation pipelines.
- Studying memory-bank compression, self-filtering and cross-dataset generalization.
- Measuring inference latency and edge-deployment trade-offs.
- Helping with the literature review and manuscript.