How embedded-machine-learning/annette works
Main Components Identification
The system should be represented with these key components:
a) Core Components:
• Estimation Module
• Benchmarking System
• Model Conversion Pipeline
• Database Storage
• Graph Processing
b) Supporting Components:
• Hardware Interfaces
• Configuration Management
• Utility Functions
[Input Networks (.pb/.onnx)] → [Model Conversion] → [ANNETTE Format]
[Estimation Pipeline]
• Mapping Models
• Layer-wise Models
[Hardware Abstraction]
• OpenVINO
• DNNDK
• NCS2
Detailed Component Breakdown
a) Estimation Module:
[Layer Model] ─── [Mapping Model]
• [Estimator] ─────┐
• [Results]
• [Benchmark Data] ────┘
b) Model Processing:
[Input Networks] → [MMDNN/ONNX Converter] → [ANNETTE Format]
• [Benchmark Runner] ───────────┘
Implementation Details
• Support for multiple frameworks (PyTorch, TensorFlow)
• Hardware acceleration support
• Benchmark data collection process
• Model optimization paths
• Result storage and analysis
Notes for Implementation
• Maintain consistent spacing and alignment
• Version information
• Brief descriptions for complex components
• Error handling paths