目录
安装python环境：
conda config --set report_errors true
conda create --name rknn3.6 python=3.6
rknn下载：
http://repo.rock-chips.com/pypi/simple/rknn-toolkit/
rknn-toolkit-1.2.1-cp36-cp36m-linux_aarch64.whl 17-Oct-2019 10:05 56659109
rknn-toolkit-1.2.1-cp36-cp36m-linux_x86_64.whl 17-Oct-2019 10:05 60841540
rknn-toolkit-1.2.1-cp36-cp36m-macosx_10_9_x86_6..> 17-Oct-2019 10:08 36891121
rknn-toolkit-1.2.1-cp36-cp36m-win_amd64.whl 17-Oct-2019 10:08 33134089
rknn-toolkit-1.2.1-cp37-cp37m-linux_aarch64.whl 17-Oct-2019 10:09 57598878
rknn-toolkit-1.3.0-cp36-cp36m-linux_aarch64.whl 20-Jan-2020 06:47 57274013
rknn-toolkit-1.3.0-cp36-cp36m-linux_x86_64.whl 20-Jan-2020 06:59 136707878
rknn-toolkit-1.3.0-cp36-cp36m-macosx_10_9_x86_6..> 20-Jan-2020 07:02 116558624
rknn-toolkit-1.3.0-cp36-cp36m-win_amd64.whl 20-Jan-2020 07:06 109971468
rknn-toolkit-1.3.0-cp37-cp37m-linux_aarch64.whl 20-Jan-2020 07:06 58575235
rknn-toolkit-1.3.2-cp37-cp37m-linux_aarch64.whl 23-Apr-2020 06:48 35742584
rknn-toolkit-1.4.0-cp36-cp36m-linux_x86_64.whl 24-Aug-2020 00:51 135521194
rknn-toolkit-1.4.0-cp36-cp36m-win_amd64.whl 24-Aug-2020 00:51 108328413
rknn-toolkit-1.4.0-cp37-cp37m-linux_aarch64.whl 24-Aug-2020 00:51 31001197
rknn-toolkit-1.4.0-cp37-cp37m-macosx_10_15_x86_..> 24-Aug-2020 00:51 108429118
rknn_toolkit-1.6.0-cp35-cp35m-linux_aarch64.whl 14-Jan-2021 02:33 34538213
rknn_toolkit-1.6.0-cp35-cp35m-linux_x86_64.whl 14-Jan-2021 02:34 160362809
rknn_toolkit-1.6.0-cp36-cp36m-linux_x86_64.whl 14-Jan-2021 02:35 160552824
rknn_toolkit-1.6.0-cp36-cp36m-macosx_10_15_x86_..> 14-Jan-2021 02:36 119510062
rknn_toolkit-1.6.0-cp36-cp36m-win_amd64.whl 14-Jan-2021 02:36 112325076
rknn_toolkit-1.6.0-cp37-cp37m-linux_aarch64.whl 14-Jan-2021 02:36 39105410
rknn_toolkit-1.6.0-cp37-cp37m-macosx_10_15_x86_..> 14-Jan-2021 02:37 117990241
rknn_toolkit-1.7.0-cp37-cp37m-linux_aarch64.whl 02-Nov-2021 08:31 40410162
rknn_toolkit-1.7.1-cp35-cp35m-linux_aarch64.whl 11-May-2022 04:33 37938676
rknn_toolkit-1.7.1-cp35-cp35m-linux_x86_64.whl 11-May-2022 04:33 142780939
rknn_toolkit-1.7.1-cp36-cp36m-linux_x86_64.whl 11-May-2022 04:34 140752080
rknn_toolkit-1.7.1-cp36-cp36m-macosx_10_15_x86_..> 11-May-2022 04:34 122255122
rknn_toolkit-1.7.1-cp36-cp36m-win_amd64.whl 11-May-2022 04:34 114232401
rknn_toolkit-1.7.1-cp37-cp37m-linux_aarch64.whl 11-May-2022 04:34 41073863
rknn_toolkit-1.7.1-cp37-cp37m-macosx_10_15_x86_..> 11-May-2022 04:35 120252888
2022.06.18更新：
rknn windows版本，只有python3.6版本，linux有3.7版本
rknn_toolkit-1.7.1-cp36-cp36m-win_amd64.whl
yolox onnx导出rknn代码
import os
import urllib
import traceback
import time
import sys
import re
import numpy as np
import cv2
from yolox.data.data_augment import preproc as preprocess
from rknn.api import RKNN
import time
from yolox.utils import mkdir, multiclass_nms, demo_postprocess, vis
from config import Config
def run():
cfg = Config()
ONNX_MODEL = 'yolox.onnx'
RKNN_MODEL = 'yolox.rknn'
rknn = RKNN(verbose=True)
rknn.list_devices()
# rknn.config(target_platform='rk3566')
rknn.config()
ret = rknn.load_onnx(model=ONNX_MODEL)
if ret < 0:
print("load faild")
ret = rknn.build(do_quantization=False,dataset="/home/bona/Projects/python/YOLOX-main/demo/RKNN/dataset.txt")
if ret < 0:
print("build faild")
ret = rknn.export_rknn(RKNN_MODEL)
if ret < 0:
print("export faild")
print("export down")