Ianzih_DeepMask

This code from foolwood/deepmask-pytorch

I made some revise.

Start

DEEPMASK=$PWD
export PYTHONPATH=$DEEPMASK:$PYTHONPATH
mkdir -p ./pretrained/deepmask   #put your pre-trained models

Testing Visualize Mask

python tools/computeProposals.py --arch DeepMask --resume $DEEPMASK/pretrained/deepmask/DeepMask.pth.tar --img ./data/test.jpg
#--resume your model , --img test img

training

1. Prepare

   mkdir -p $DEEPMASK/data/coco ; cd $DEEPMASK/data/coco
   mkdir train ;  mkdir val ; #train dir. put your train_img , val dir. put your val_img
   mkdir annotations ; #put your train & val .json

2. traing

   CUDA_VISIBLE_DEVICES=0,1,2,3 python tools/train.py --dataset coco -j 20 --freeze_bn