Getting Started

we visualize our training details via wandb (https://wandb.ai/site).

visualization

1. you'll need to login

   $ wandb login
   

2. you'll need to copy & paste you API key in terminal

   $ https://wandb.ai/authorize
   

   or add the key to the "code/config/config.py" with

   C.wandb_key = ""
   

training

our code is trained using one nvidia A6000, but our code also supports distributed data parallel mode in pytorch. We set batch_size=8 for all the experiments, with learning rate 7.5e-6 and 700 * 700 resolution.

checkpoints

we follow Meta-OoD and use the deeplabv3+ checkpoint in here. you'll need to put it in "ckpts/pretrained_ckpts" directory, and please note that downloading the checkpoint before running the code is necessary for our approach.

for training, simply execute

$ python rpl_corocl.code/main.py 

inference

please download our checkpoint from here and specify the checkpoint path for rpl_corocl_weight_path in config.py.

python rpl_corocl.code/test.py