代码拉取完成，页面将自动刷新
Just follow the ncnn official tutoral of build-for-linux to install ncnn. Following steps are all carried out on my raspberry pi:
step 1: install dependencies
$ sudo apt install build-essential git cmake libprotobuf-dev protobuf-compiler libopencv-dev
step 2: (optional) install vulkan
step 3: build
I am using commit 6869c81ed3e7170dc0, and I have not tested over other commits.
$ git clone https://github.com/Tencent/ncnn.git
$ cd ncnn
$ git reset --hard 6869c81ed3e7170dc0
$ git submodule update --init
$ mkdir -p build
$ cmake -DCMAKE_BUILD_TYPE=Release -DNCNN_VULKAN=OFF -DNCNN_BUILD_TOOLS=ON -DCMAKE_TOOLCHAIN_FILE=../toolchains/pi3.toolchain.cmake ..
$ make -j2
$ make install
$ python -m pip install onnx-simplifier
On your training platform:
$ cd BiSeNet/
$ python tools/export_onnx.py --aux-mode eval --config configs/bisenetv2_city.py --weight-path /path/to/your/model.pth --outpath ./model_v2.onnx
$ python -m onnxsim model_v2.onnx model_v2_sim.onnx
Then copy your model_v2_sim.onnx from training platform to raspberry device.
On raspberry device:
$ /path/to/ncnn/build/tools/onnx/onnx2ncnn model_v2_sim.onnx model_v2_sim.param model_v2_sim.bin
You can optimize the ncnn model by fusing the layers and save the weights with fp16 datatype.
On raspberry device:
$ /path/to/ncnn/build/tools/ncnnoptimize model_v2_sim.param model_v2_sim.bin model_v2_sim_opt.param model_v2_sim_opt.bin 65536
$ mv model_v2_sim_opt.param model_v2_sim.param
$ mv model_v2_sim_opt.bin model_v2_sim.bin
You can also quantize the model for int8 inference, following this tutorial. Make sure your device support int8 inference.
On raspberry device:
$ mkdir -p BiSeNet/ncnn/build
$ cd BiSeNet/ncnn/build
$ cmake .. -DNCNN_ROOT=/path/to/ncnn/build/install
$ make
./segment
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。
马建仓 AI 助手