Code for "Candidate-aware Selective Disambiguation Based On Normalized Entropy for Instance-dependent Partial-label Learning" (ICCV 2023).
Requirements:
python 3.9.12
torch 1.12.1
torchvision 0.15.2
Dataset:
./"dataset_root"
Candidate label generation:
Model checkpoints saved in ./model_path
Run:
sh run.sh 0 cifar10 10
sh run.sh 1 cifar100 100
sh run.sh 2 fmnist 10
sh run.sh 3 kmnist 10
Parameters varied in run.sh.
Log file saved in ./output_log