This is the foreground segmentation task with semi-supervised learning.
You can download the dataset here:
CDNet2014
SBI
SBM-RGBD
The dataset should be organized in the following way:
./dataset/
/label/
/category1/
/input/...
/groundtruth/...
/background/...
/category2/...
/unlabel/
/category1/
/input/...
/groundtruth/
/background/...
/category2/...
You can train the model with the following command.
More options can be found in "./options/BaseOptions.py".
python train.py --batchSize 4 --lr 0.0001 --cuda True
You can view the results of the model with the following command.
Here are some sample results:
cd ./utils
python visualize.py