Awesome

Darknet Logo

./darknet detector train cfg/wildhog.data cfg/yolov3-tiny.cfg yolov3-tiny.conv.15 -gpus 0 >> backup_wildhog/wildhog_yolov3_tiny.txt

Features of this Darknet

In tools/, some tools for darknet.
In api/, a API interface for darknet.

Quick run the pretrained weights

Run on one Image

./darknet detector test cfg/coco.data cfg/yolov3.cfg weights/yolov3.weights ./data/dog.jpg

Run on a series of images

./darknet detector list cfg/coco.data cfg/yolov3.cfg weights/yolov3.weights /media/elab/sdd/data/TinyTLP/Sam/img/

Reat-Time Detection on a Webcam

./darknet detector demo cfg/coco.data cfg/yolov3.cfg weights/yolov3.weights

Train with your own data with reference [Link]

Train Yolov3 without pre-trained weights:

./darknet detector train cfg/wildhog.data cfg/wildhog-yolov3-tiny.cfg -gpus 0

Train Yolov3 with pre-trained weights:

./darknet detector train cfg/wildhog.data cfg/wildhog-yolov3-tiny.cfg darknet53.conv.74 -gpus 0 >> wildhog_yolov3_tiny.txt

Train Yolov3-tiny with pre-trained weights:

First prepare the pretrained weights:

./darknet partial cfg/yolov3-tiny.cfg yolov3-tiny.weights yolov3-tiny.conv.15 15

Train with the command:

./darknet detector train cfg/wildhog.data cfg/wildhog-yolov3-tiny.cfg  yolov3-tiny.conv.15 -gpus 0

After train, run with:

./darknet detector test cfg/wildhog.data cfg/wildhog-yolov3-tiny.cfg backup_wildhog/wildhog-yolov3-tiny_final.weights ./wildhog.jpg

./darknet detector test cfg/wildhog.data cfg/wildhog-yolov3-tiny.cfg backup_wildhog/wildhog-yolov3-tiny.backup 

./darknet detector demo cfg/wildhog.data cfg/wildhog-yolov3-tiny.cfg backup_wildhog/wildhog-yolov3-tiny.backup

Train Imagenet

Download the data and create the list file and put them in the data/ folder:

find `pwd`/ILSVRC2012_img_train -name \*.JPEG > imagenet1k.train.list
find `pwd`/ILSVRC2012_img_val -name \*.JPEG > imagenet1k.valid.list

Training coco and voc

Remember to create backup_** before training

./darknet detector train cfg/coco.data cfg/yolov3.cfg darknet53.conv.74 -gpus 1
./darknet detector train cfg/voc.data cfg/yolov3-voc.cfg darknet53.conv.74 -gpus 0

Run on Mac

brew install opencv@2
echo 'export PATH="/usr/local/opt/opencv@2/bin:$PATH"' >> ~/.bash_profile

each time to open a new terminal, you should run these to make:

export LDFLAGS=-L/usr/local/opt/opencv@2/lib
export CPPFLAGS=-I/usr/local/opt/opencv@2/include
export PKG_CONFIG_PATH=/usr/local/opt/opencv@2/lib/pkgconfig

Net compression (not finished)

./darknet detector test cfg/coco.data cfg/yolov3.cfg weights/yolov3.weights data/dog.jpg

GPU=0 CUDNN=0 OPENMP=0: 21.315930 seconds.

GPU=0 CUDNN=0 OPENMP=1: 4.199370 seconds.

yolov3.cfg: width=208 height=208: 2.430844 seconds.

Run with OpenTracker

Copy trackers' source file from [OpenTracker] and make install:

git clone https://github.com/rockkingjy/OpenTracker.git
cd OpenTracker
make -j`nproc`
sudo make install

set flag in makefile:

OPENTRACKER=1

and run:

./darknet detector tracking cfg/coco.data cfg/yolov3.cfg weights/yolov3.weights

It will first run the yolo detection. Then when a person comes into the vision field, then it will automatically do the tracking.

Run with a touchscreen and OpenTracker:

In makefile set:

TS=1

Assuming that your touchscreen is connected to /dev/input/event6, else modify it (must be run under sudo):

sudo ./darknet detector tracking cfg/coco.data cfg/yolov2-tiny.cfg yolov2-tiny.weights "/dev/input/event6"

sudo ./darknet detector tracking cfg/coco.data cfg/yolov3.cfg yolov3.weights "/dev/input/event4"

It will first run the yolo detection, with touchscreen, you choose the object to track, then it will track automatically.

Run with maestro motor

In Makefile set:

MAESTRO=1