v2.0.0 #17

AnomaVision - Multi-Format Anomaly Detection Inference

A production-ready anomaly detection system with comprehensive model export and inference capabilities.

Key Features:

• 🔄 Multi-format export: Convert PyTorch models to ONNX, TorchScript, OpenVINO
• 🚀 Unified inference: Single API for all model formats (.pt, .onnx, .torchscript, OpenVINO dirs)
• 🎯 Auto-detection: Automatic backend selection based on file extension
• ⚡ Optimized backends: ONNX Runtime, TorchScript JIT, Intel OpenVINO
• 🖥️ Cross-platform: CPU/GPU support with device-specific optimizations

Quick Start:

# Export model to all formats
python export.py --model padim_model.pt --format all
# Run inference (auto-detects format)
Usage - formats:
$ python detect.py --model padim_model.pt # PyTorch
 padim_model.torchscript # TorchScript
 padim_model.onnx # ONNX Runtime
 padim_model_openvino # OpenVINO
 padim_model.engine # TensorRT

Supported Formats:

• PyTorch (.pt) - Development & training
• TorchScript (.torchscript) - Production deployment
• ONNX (.onnx) - Cross-platform compatibility
• OpenVINO (dirs/.xml) - Intel hardware optimization

Perfect for deploying anomaly detection models across different environments and hardware platforms.