taloot

> Hi, @taloot thanks for your issue. > > * what is the inference time of yolov5s at 512x512 with Khadas VIM3 NPU? > * some type op in yolov5s...

> Hi, @taloot thanks for your issue. > > * what is the inference time of yolov5s at 512x512 with Khadas VIM3 NPU? > * some type op in yolov5s...

the vim3 didnt work for me at all coral tpu works and every day it become better me i thought the npu performance will be better, but i found out...

https://www.youtube.com/watch?v=D9IExho8pwo this is updated link

The problem never happend on 416 It happend if i want more than 416 I m using currently 7 classes and 512 I wish i can run more than 512...

https://www.sendgb.com/upload/?utm_source=iRw3E1XMmHY this is the files u may want, when i make the freeze graphic i have to specify the resloution if i have more than 7 classes the compile stops...

[161Newds-int8_edgetpu.log](https://github.com/google-coral/edgetpu/files/6722344/161Newds-int8_edgetpu.log) Edge TPU Compiler version 15.0.340273435 Input: 161Newds-int8.tflite Output: 161Newds-int8_edgetpu.tflite Operator Count Status CONV_2D 86 Mapped to Edge TPU QUANTIZE 24 Mapped to Edge TPU MAX_POOL_2D 3 Mapped to Edge...

> Hi @taloot, I was able to run 416 model on 98 classes without an issue. Would you mind sharing your postquantized tflite and edgetpu model ? I can maybe...

`import argparse import logging import os import sys import traceback from copy import deepcopy from pathlib import Path sys.path.append('./') # to run '$ python *.py' files in subdirectories import numpy...

[161512-int8_edgetpu.zip](https://github.com/google-coral/edgetpu/files/6729632/161512-int8_edgetpu.zip) this is my working high speed yolov5s@512 this one im able to reach 20fps solid yolov5s 512x512 = 45ms yolov5s 448x448 = 38ms yolov5s 416x416=26ms yolov5s 320x320 = 18ms...