同步操作将从 hong_liu/sophon-inference 强制同步,此操作会覆盖自 Fork 仓库以来所做的任何修改,且无法恢复!!!
确定后同步将在后台操作,完成时将刷新页面,请耐心等待。
_____ __ ____ ____ / ___/____ ____ / /_ ____ ____ / _/___ / __/__ ________ ____ ________ \__ \/ __ \/ __ \/ __ \/ __ \/ __ \______ / // __ \/ /_/ _ \/ ___/ _ \/ __ \/ ___/ _ \ ___/ / /_/ / /_/ / / / / /_/ / / / /_____// // / / / __/ __/ / / __/ / / / /__/ __/ /____/\____/ .___/_/ /_/\____/_/ /_/ /___/_/ /_/_/ \___/_/ \___/_/ /_/\___/\___/ /_/
Guide to deploying deep-learning inference networks and deep vision primitives on Sophon TPU.
## Prerequisites and Compilation
- **BMNNSDK2** Required
- **CMake** Required
- **OpenCV3(at least 3.4.6)** Required for C++ samples
- **Python3** Optional for python samples
- **Sphinx** Optional for documents
```shell
# install bmnnsdk2
cd bmnnsdk2/scripts
# extract libs adaptable to the OS
./install_lib.sh nntc
# remove old driver and install the new one, 'pcie' for x86_64 and 'arm_pcie' for aarch64
sudo ./remove_driver_pcie.sh # for x86_64
sudo ./uninstall_driver_pcie.sh # for x86_64
# configure the environment. 'pcie' for x86_64 and 'arm_pcie' for aarch64
source ./envsetup_pcie.sh # for x86_64
# compilation
mkdir build
cd build
cmake ..
make
ID | Input | Decoder | Preprocessor | Data Type | Model | Mode | Model Number | TPU Number | Multi-Thread |
---|---|---|---|---|---|---|---|---|---|
1 | image | opencv | opencv | fp32/int8 | resnet-50 | static | 1 | 1 | N |
2 | image | opencv | opencv | fp32/int8 | resnet-50 | static | 1 | 1 | Y |
3 | image | opencv | opencv | fp32/int8 | resnet-50 | static | 1 | 2 | Y |
4 | image | opencv | opencv | fp32/int8 | resnet-50 | static | 2 | 1 | Y |
ID | Input | Decoder | Preprocessor | Data Type | Model | Mode | Batch Size | Multi-Thread |
---|---|---|---|---|---|---|---|---|
1 | video/image | opencv | opencv | fp32/int8 | ssd_vgg | static | 1 | N |
2 | video/image | bm-ffmpeg | bmcv | fp32/int8 | ssd_vgg | static | 1 | N |
3 | video | bm-ffmpeg | bmcv | fp32/int8 | ssd_vgg | static | 4 | N |
4 | video/image | bm-opencv | bm-opencv | fp32/int8 | ssd_vgg | static | 1 | N |
5 | video/image | bm-opencv | bmcv | fp32/int8 | ssd_vgg | static | 1 | N |
6 | video | opencv | opencv | fp32/int8 | yolov3 | static | 1 | Y |
7 | video | bm-ffmpeg | bmcv | fp32/int8 | yolov3 | static | 1 | Y |
ID | Input | Decoder | Preprocessor | Data Type | Model | Mode |
---|---|---|---|---|---|---|
1 | image | opencv | opencv | fp32 | mtcnn | dynamic |
C++ Samples Usage Instructions
Python Samples Usage Instructions
Result files of each examples to compare are in directories: ./release/qa_test/sc5_tests/cpp_OR_python/CASE_NAME/verify_files/
Please read CONTRIBUTING.md for details on our code of conduct, and the process for submitting pull requests to us.
See also the list of Contributors who participated in this project.
This project contains codes written in C++, Python and Shell. We refer to Google Style Guides with some minor modifications.
This project is licensed under the Apache License, Version 2.0.
此处可能存在不合适展示的内容,页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容,可点击提交进行申诉,我们将尽快为您处理。