guillherms/vision-architecture-analyzer-yolo11-detect

Vision Architecture Analyzer YOLO11 Detect v1.0.0

This model was trained with YOLO11 for detecting components in cloud architecture diagrams (AWS and Azure).

The goal is to automatically identify architecture nodes such as compute, database, cache, storage, gateways, boundaries, and external services from images of technical diagrams.

Dataset

• Architecture diagrams manually annotated in Label Studio
• Classes include actors, services, data, observability, and boundaries
• Bounding boxes in YOLO format (detection only)

Training Configuration

Parameter	Value
Base Model	YOLO11s
Image Dimension	1280x1280
Augmentations	Disabled (mosaic, mixup, flip, perspective)
Focus	Icons and architectural structures

Note: Augmentations were disabled to preserve the geometry of the diagrams.

Usage

The model can be used to automatically extract components from a diagram and feed pipelines for:

• Threat Modeling (STRIDE)
• Architecture Analysis
• Automatic Security Documentation Generation

Input and Output

Type	Description
Input	Architecture diagram image
Output	Bounding boxes with detected classes and confidence scores

Version

• v1.0.0 — First stable model for architecture node detection