Anish Kumar Chada@Anish_1207

VisionInspect AI — Industrial Visual Defect Detection With Limited Training Data The Challenge: In high-throughput manufacturing (electronics, bottling, metal fabrication, textiles), defects are exceptionally rare (occurring in less than 0.1% of parts), yet shipping a defective unit can lead to catastrophic recalls. Conventional supervised deep learning models fail in this environment because they require thousands of annotated defect examples for every failure mode and are fundamentally blind to novel, unforeseen defect types that emerge during production. Our Solution: VisionInspect AI is an end-to-end industrial visual inspection system that solves defect scarcity by learning exclusively from normal (defect-free) parts. Using advanced feature-embedding memory banks (PatchCore) and spatial Gaussian distribution modeling (PaDiM), the system models the nominal appearance manifold of products without ever seeing a defect during training. Any structural, textural, or geometric deviation in incoming parts is immediately detected as an anomaly. Key Innovations & Capabilities: Zero-Defect Prior Training: Reaches state-of-the-art defect detection (98.2% Image AUROC on MVTec benchmarks) using only nominal golden samples, drastically outperforming few-shot supervised classifiers on novel flaws. Pixel-Level Localization: Computes Gaussian-smoothed anomaly heatmaps and extracts CAD-ready bounding boxes, centroids, and defect surface-area percentages in real time. Conformal Uncertainty Routing: Replaces rigid single thresholds with statistical prediction intervals (NORMAL → → line pass, DEFECT → → scrap ejection, BORDERLINE → → operator review), bounding false alarms and eliminating escape risk. Enterprise Dashboard: Built on an asynchronous FastAPI backend and a modern React 19 operator UI supporting real-time single-image analysis, batch directory processing, and an interactive human review triage queue.

Six Matches. One leaderboard. Build on AMD, earn points and compete for prizes.

🌎 Hybrid Hackathon | 💻 Online Build: 12-17 October 2026 | On-site phase: 17-18 October 2026 To participate in the hackathon, simply click Sign up with AMD. If you are not already a member of the AMD AI Developer Program (ADP), you will need to create an account. The AMD Developer Hackathon: ACT III is a hands-on event for developers, founders, engineers and builders who want to push what’s possible with AI on real infrastructure. 📍 If approved for the on-site experience, you'll also have the opportunity to showcase your project live. Travel and accommodation will not be covered.