Machine Learning•2025
Visual Anomaly Detection System
Convolutional model trained to detect structural surface variations.
[01] The Problem
Technical Constraints & Motivation
Manual inspection across repetitive image datasets is slow and error-prone, requiring an automated classification model.
[02] What I Built
Architecture & Implementation
Computer vision classification workflow including dataset augmentation, transfer learning fine-tuning, and export for local inference.
- —Dataset pipeline with data augmentations to mitigate overfitting
- —Exported model weights to ONNX format for cross-platform inference
- —Evaluation dashboard displaying confusion matrices and per-class precision
[03] Outcome & Results
Delivered System
Trained a convolutional classifier on annotated image samples with class-weighted loss to handle imbalanced categories.
System Interface & Wireframes1 Image
Visual Anomaly Detection model evaluation and heatmap output[Placeholder Preview — Real Screenshot to be Added]