MLPerf benchmark automation demo¶
We collaborate with the MLPerf benchmarking consortium to automate the manual and tedious process of creating MLPerf inference benchmarks and submitting results.
We have prepared a demo of the open-source CK solution for the object detection application from the MLPerf inference benchmark based on SSD-Mobilenets, TensorFlow and COCO data-set.
Please follow the instructions from above page to install the open-source client with the unified API to automatically build, run and validate this solution on your machine.
You should be able to reproduce some of the latest MLPerf inference benchmark results from this research paper, participate in crowd-benchmarking and compare your results against the official ones using this live SOTA scoreboard.
You can also check the live demo of this solution using the webcam in your browser: