Darknet with NNPACK

NNPACK was used to optimize Darknet without using a GPU. It is useful for embedded devices using ARM CPUs.

Build from Raspberry Pi 3

Log in to Raspberry Pi using SSH.
Install PeachPy

git clone https://github.com/Maratyszcza/PeachPy.git
cd PeachPy
python setup.py build
sudo python setup.py install

Install Ninja

git clone https://github.com/ninja-build/ninja.git
cd ninja
git checkout release
./configure.py --bootstrap

Install clang

sudo apt-get install clang

Install NNPACK-rpi

git clone https://github.com/thomaspark-pkj/NNPACK-rpi.git --recursive
cd NNPACK-rpi
git checkout raspberrypi
python ./configure.py --enable-rpi --enable-shared
$NINJA_PATH/ninja
sudo cp -a lib/libnnpack.* /usr/lib/
sudo cp include/nnpack.h /usr/include/
sudo cp third-party/pthreadpool/include/pthreadpool.h /usr/include/

Build darknet-nnpack

git clone https://github.com/thomaspark-pkj/darknet-nnpack.git
git checkout nnpack
make

Test

The weight files can be downloaded from the YOLO homepage.

YOLO
./darknet detector test cfg/coco.data cfg/yolo.cfg yolo.weights data/person.jpg
Tiny-YOLO
./darknet detector test cfg/coco.data cfg/tiny-yolo.cfg tiny-yolo.weights data/person.jpg

Result

Model	Build Options	Prediction Time (seconds)
YOLO	NNPACK=1,ARM_NEON=1	7.37
YOLO	NNPACK=0,ARM_NEON=0	185
Tiny-YOLO	NNPACK=1,ARM_NEON=1	1.78
Tiny-YOLO	NNPACK=0,ARM_NEON=0	38