wangmaolin / lpcv2020-hku

Volunteer to search "good" network architectures for Xilinx DPUs

LPCV2020 Contest

Board Setup

1. flash PYNQ V2.5 to the SD card
2. upgrade the board using https://github.com/Xilinx/DPU-PYNQ

Test Different Models

1. Rename the util-modelXX.py to utils.py and put it on the board together with model file: modelXX.elf.
2. Run test_accuracy_latency.ipynb to get the latency and accuracy of the tested model.
3. Change the threadnum in utils.py to use different number of threads.

Image files

The ImageNet validation set should be put under ./val which contains ILSVRC2012_val_00000001.JPEG, ILSVRC2012_val_00000002.JPEG, ..., ILSVRC2012_val_00050000.JPEG.

Note

All the model weights and activations are quantized to 8 bits.