Probabilistic Domain Adaptation on LIVECell Dataset

Dataset is publicly available here

Implements the domain adaptation experiments on the LIVECell dataset. Each script implements training, prediction and validation of the respective model. All the required data can be downloaded with the script prepare_data.py.

For more information on the individual scripts refer to their help output, e.g. python livecell_unet.py -h.

livecell_unet.py (Source UNet Training)

python livecell_unet.py [--train / --predict / --evaluate]
                        --data <PATH-TO-LIVECELL-DATA>
                        [(optional) --pred_path <PATH-TO-SAVE-PREDICTIONS>]

livecell_punet.py (Source PUNet Training)

python livecell_punet.py [--train / --predict / --evaluate]
                         --data <PATH-TO-LIVECELL-DATA>
                         [(optional) --pred_path <PATH-TO-SAVE-PREDICTIONS>]

livecell_punet_target.py (Target PUNet Training using Pseudo Labels from Source)

python livecell_punet_target.py --get_pseudo_labels
                                [--train / --predict / --evaluate]
                                [(optional : allows consensus masking) --consensus]
                                --data <PATH-TO-LIVECELL-DATA>
                                [(optional) --pred_path <PATH-TO-SAVE-PREDICTIONS>]

livecell_mt.py (Mean-Teacher Separate Training)

python livecell_mt.py [--train / --predict / --evaluate]
                      [(optional : enables consensus weighting) --consensus]
                      [(optional : enables consensus masking) --consensus --masking]
                      --data <PATH-TO-LIVECELL-DATA>
                      [(optional) --pred_path <PATH-TO-SAVE-PREDICTIONS>]

livecell_fm.py (FixMatch Separate Training)

python livecell_fm.py [--train / --predict / --evaluate]
                      [(optional : enables consensus weighting) --consensus]
                      [(optional : enables consensus masking) --consensus --masking]
                      --data <PATH-TO-LIVECELL-DATA>
                      [(optional) --pred_path <PATH-TO-SAVE-PREDICTIONS>]

livecell_adamatch.py (FixMatch-based Joint Training)

python livecell_adamatch.py [--train / --predict / --evaluate]
                            [(optional : enables consensus weighting) --consensus]
                            [(optional : enables consensus masking) --consensus --masking]
                            --data <PATH-TO-LIVECELL-DATA>
                            [(optional) --pred_path <PATH-TO-SAVE-PREDICTIONS>]

livecell_adamt.py (Mean-Teacher-based Joint Training)

python livecell_adamt.py [--train / --predict / --evaluate]
                         [(optional : enables consensus weighting) --consensus]
                         [(optional : enables consensus masking) --consensus --masking]
                         --data <PATH-TO-LIVECELL-DATA>
                         [(optional) --pred_path <PATH-TO-SAVE-PREDICTIONS>]