This repository forks from darknet-nnpack and adds support for DREML functions. These functions are enabled by macros in darknet.h. The macros and their usecases are explained below:
- CUSTOM_BACKPROP: Enables custom loss functions in backpropogation of region_layer, yolo_layer and softmax_layer
- DYNAMIC_FMAP_PRUNING: Prevents backpropogation of convolutional layers from calculating weight updates
- IMG_SEG: Adds support for image segmentation as proposed here, should be undefined for other applications
- DET_THRESH: Detection threshold used by custom loss functions of region_layer and yolo_layer
If you have a question about this repository or a problem using modified darknet, you can contact Kamyar at [email protected]. Note that, since Confu support for NNPACK is now deprecated, you can build this fork of NNPACK with CMake. The remainder of this README is same as the original repo.
NNPACK was used to optimize AlexeyAB/darknet without using a GPU. It is useful for embedded devices using ARM CPUs.
Log in to Raspberry Pi using SSH.
Install PeachPy and confu
sudo pip install --upgrade git+https://github.com/Maratyszcza/PeachPy
sudo pip install --upgrade git+https://github.com/Maratyszcza/confu
Install Ninja
git clone https://github.com/ninja-build/ninja.git
cd ninja
git checkout release
./configure.py --bootstrap
export NINJA_PATH=$PWD
Install clang
sudo apt-get install clang
Install NNPACK-darknet
git clone https://github.com/digitalbrain79/NNPACK-darknet.git
cd NNPACK-darknet
confu setup
python ./configure.py --backend auto
$NINJA_PATH/ninja
sudo cp -a lib/* /usr/lib/
sudo cp include/nnpack.h /usr/include/
sudo cp deps/pthreadpool/include/pthreadpool.h /usr/include/
Build darknet-nnpack
git clone https://github.com/digitalbrain79/darknet-nnpack.git
cd darknet-nnpack
make
COCO trained weights files can be downloaded from the AlexeyAB/darknet.
COCO
./darknet detector test cfg/coco.data [cfg file] [weights file] [image path]
Pascal VOC
./darknet detector test cfg/voc.data [cfg file] [weights file] [image path]
| cfg | Build Options | mAP | Prediction Time (seconds) |
|---|---|---|---|
| yolov3-tiny.cfg | NNPACK=1 | 33.1 | 1.1 |
| yolov3-tiny.cfg | NNPACK=0 | 14.5 | |
| yolov3-tiny-prn.cfg | NNPACK=1 | 33.1 | 0.86 |
| yolov3-tiny-prn.cfg | NNPACK=0 | 9.3 |
| cfg | Build Options | mAP | Prediction Time (seconds) | Weights file |
|---|---|---|---|---|
| yolov3-tiny-voc.cfg | NNPACK=1 | 65.9 | 1.0 | yolov3-tiny-voc.weights |
| yolov3-tiny-voc.cfg | NNPACK=0 | 14.0 | ||
| yolov3-tiny-prn-voc.cfg | NNPACK=1 | 65.2 | 0.77 | yolov3-tiny-prn-voc.weights |
| yolov3-tiny-prn-voc.cfg | NNPACK=0 | 8.9 | ||
| Gaussian_yolov3-tiny-voc.cfg | NNPACK=1 | 65.7 | 1.0 | Gaussian_yolov3-tiny-voc.weights |
Download OS image from here
sudo dd bs=4M if=darknet-nnpack.img of=/dev/sdX conv=fsyn