6.12. Performing Inference on YOLOv3 and Calculating Accuracy Metrics
- You have a copy of of OpenVINO™ Model Zoo 2022.3.1 in your $COREDLA_ROOT/demo/open_model_zoo/ directory.
- You have completed the steps in Preparing OpenVINO Model Zoo.
- You have programmed the Intel® Arria® 10 FPGA device with a bitstream that corresponds to A10_Performance.arch. For example instructions, refer to Building an FPGA Bitstream for the PCIe Example Design.
To target this tutorial to a different FPGA device, you can adjust the commands appropriately (such as using a a different architecture file like AGX7_Performance.arch to target Intel Agilex® 7 devices). Be sure that you are using an architecture file that includes support for YOLOv3 (and specifically the PReLU activation).
The steps in the sections that follow guide you through preparing a YOLOv3 model for inference, preparing a COCO validation dataset and annotations, as well as calculating mAP and COCO AP metrics.