Single Shot MultiBox Detector implement in PyTorch and OpenCV.
python >= 3.5.0
pip3 package
torch >= 0.4.0 # or more
opencv >= 3.3 # or more
mkdir model
cd model
wget https://download.pytorch.org/models/vgg16-397923af.pth
python3 script/convert_vgg.py
From here VOC12 download train data and unzip to data folder like:
$REPO_PATH/data/VOC2012_trainval_train_images/
recommand use pretrain model, if not, the loss desent slowly
src/train.py [--lr] [--epoch] [--resume]
src/test.py [img_path]
This only trained by voc dataset with 7 epoch
This implementation is initially inspired by: