The following are the current general sizes (MB) of AutoML Vision Edge models. These values are based on the model optimization criteria and model type:
Model optimized for | Format | Model size |
---|---|---|
High accuracy | Container, TensorFlow.js (tfjs), CoreML | 23 MB |
TFlite | 6 MB | |
Best tradeoff | TensorFlow, TensorFlow.js (tfjs), CoreML | 12.4 MB |
TFlite | 3.2 MB | |
Low latency | TensorFlow, TensorFlow.js (tfjs), CoreML | 2 MB |
TFlite | 0.6 MB |