This README would normally document whatever steps are necessary to get your application up and running.
git clone git@bitbucket.org:toancauxanh/yolo-aitl.git
cd yolo-aitl
python3 -m venv --system-site-packages venv
source venv/bin/activate
(venv) pip install -r requirements.txt
(venv) ./init.py
For simpler, the commands below will be presented without the prefix "(venv)".
Depend on your system, let's choose tensorflow or tensorflow-gpu should be installed:
pip install tensorflow
# or
pip install tensorflow-gpu
Then:
- download yolo.weights into
temp/weights/ - download initial-model.h5 into
temp/checkpoints/ - download person-small.tar.gz (10MB) or person-full.tar.gz (310MB)
- extract compressed
.tar.gzfile intotemp/folder - run
seedto generate training set & evaluation set
For example:
cd yolo-aitl
tar -xzvf temp/person-small.tar.gz -C temp
./seed.py -d temp/person-small/
seed.py would automatically separate dataset to training set and evaluation set for you.
Lastly, if everything is ok, you can start training now:
./train.py
Using predict.py, you can:
1, Detect person from specified image
./predict.py -f tests/images/01.jpg
2, Detect all .jpg image from specified folder
./predict.py -d path_to_input_folder -o path_to_output_folder
3, Randomly detect an image stored in test/images
./predict.py
4, Detect person from AVI or MP4 video file
./predict.py -f path_to_your_video.avi
5, Detect persom from camera
./predict -c N
With N is camera index.
tensorboar --logdir=temp/logs
./test.sh