Hi, I have deployed all the convolution layers of YOLOv3(256x256) to VTA(200 MHz) on ZCU104, through convolution scheduling optimization, The final model recognition speed is about 3.4 FPS.
The results are as follows:
Hi, I have deployed all the convolution layers of YOLOv3(256x256) to VTA(200 MHz) on ZCU104, through convolution scheduling optimization, The final model recognition speed is about 3.4 FPS.
The results are as follows:
Hi @xjtu_soc May I ask what version of the code you used? Can you share the implementation details of the conv2d scheduling?Thank you.