Models
Available pre-built models, dataset adapters for accuracy measurement, and additional verified architectures.
| Page | What it covers |
|---|---|
| Model Zoo | All pre-built models included with the SDK: object detection, classification, tracking, segmentation, keypoint detection, and language models. Includes task categories, input sizes, and accuracy benchmarks. |
| Model Recipes | Standalone Python recipes for selected Model Zoo models, with readable compile and development-check paths. |
| Additional Models | Classification models verified on Metis but not yet in the Model Zoo. Includes accuracy drop figures versus FP32 baseline. |
| Dataset Adapters | YAML configuration reference for dataset adapters used in accuracy measurement and custom deployment. Covers ObjDataAdapter, TorchvisionDataAdapter, and their variants. |