Fix SurfaceAI models to use EfficientNetV2 architecture
rezzzq+1: Claude
bb548d7
- SurfaceAI models are EfficientNetV2-S, not YOLO
- Added EfficientNetClassifier wrapper for loading and inference
- Quality models are regression (single output), others are classification
- Added torch and torchvision to requirements
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Add multi-model inference handler with RDD and SurfaceAI support
rezzzq+1: Claude
3138c42
- business/finishing: classification models for road assessment
- rdd: road damage detection with bounding boxes (YOLO12s RDD2022)
- surfaceai: surface type, road type, and quality classification
- Automatically selects quality model based on detected surface type
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
