您好，我在pynq上跑了您的yolo程序，有些问题想请教一下您 about yolov2_xilinx_fpga

[wxbbuaa2011]

very good work！

[Super-Alpha]

请问一下，vivado hls 的TestBench Files 是在那一部分当中？

[dhm2013724]

请问一下，vivado hls 的TestBench Files 是在那一部分当中？

I'm writing the accelerator of YOLOv2-Tiny, maybe you can be waiting until Friday this week? I will add the testbench of YOLOv2 and add the accelerator of YOLOv2-Tiny to this project.
Thanks!

[Super-Alpha]

Thank you ! I am willing to wait!

[dhm2013724]

Thank you ! I am willing to wait!

I have added the testbench to hls repo, but it will spend too much time. GOOD LUCK!

[wxbbuaa2011]

Thank you， I want to follow your work.

[dhm2013724]

Hello, I trained the network with my own data set. How can I convert the trained parameters into your format, can you help me?

Just use the traditional convert method in reference[6]. If you can wait for two weeks, the whole design will be available. I will create anthor project to introduce the design in detail. But I will use C/C++ source code in SDK instead of Python. The design is still being optimized, maybe the latency will decrease to nearly 1s.

[TriggerSignal]

I saw in the yolov2_hls_ps function in the yolov2.h file that there is an operation to read the yolov2_w_reorg_bn_ap16_short16.bin file. However, there is no such file in the project. Where is this file? What is the content inside?

[TriggerSignal]

I saw in the yolov2_hls_ps function in the yolov2.h file that there is an operation to read the yolov2_w_reorg_bn_ap16_short16.bin file. However, there is no such file in the project. Where is this file? What is the content inside?

You can find it in the HLS folder and have a link

Thank you very much, have found, the simulation has passed

[dhm2013724]

Hello, I trained the network with my own data set. How can I convert the trained parameters into your format, can you help me?

Just use the traditional convert method in reference[6]. If you can wait for two weeks, the whole design will be available. I will create anthor project to introduce the design in detail. But I will use C/C++ source code in SDK instead of Python. The design is still being optimized, maybe the latency will decrease to nearly 1s.

First of all, thank you very much for sharing. I have a question to ask. I have a new yolov2.weights. How can I convert it into weight.weights and bias.weights? I need to convert the model to caffemodel first. Follow-up work? If not, can you provide a reference method for you? thank you very much

Hi, I'm meeting a deadline, but I will create a new project to explain it in two days. Related link will be available in the homepage. Good Luck!