MLPerf Automotive is a benchmark suite for measuring how fast automotive systems can run models in a variety of deployment scenarios.
Use the r0.5 branch (git checkout v0.5abtf) if you want to reproduce v0.5 results.
See the individual Readme files in the reference app for details.
| model | reference app | framework | dataset |
|---|---|---|---|
| ssd-resnet50 | v0.5/2D object detection | pytorch, onnx | Cognata |
| bevformer-tiny | v0.5/camera-based 3D object detection | pytorch, onnx | NuScenes |
| DeepLabV3+ | v0.5/semantic segmentation | pytorch, onnx | Cognata |