Borrowing from yourself: Faster future video segmentation with partial channel update
pip install -r requirements.txt
- Choose model and configuration from
scripts
folder - Add path to the root of the Cityscapes dataset in the json config file in variable
db_root
- Run training using the
run.sh
script with the folder name as first parameter and the config file name as second parameter
Examples:
./run.sh deeplab_base config_50.json
./run.sh swiftnet_base config.json
- Check model and configuration of trained model from
scripts
folder - Run evaluation using the
run.sh
script with the folder name as first parameter and the config file name as second parameter, and the path to the learned model as third parameter
Example:
./run.sh deeplab_base config_50.json runs/deeplab/dl_50_sgd/experiment_0/checkpoint.pth