search

Alibaba-MIIL / PartialLabelingCSL

Official implementation for the paper: "Multi-label Classification with Partial Annotations using Class-aware Selective Loss"
MIT License
129 stars 18 forks source link

How can I apply it to custom dataset? #11

Closed wasdzzzzzz closed 2 years ago

wasdzzzzzz commented 2 years ago

--prior_path=./outputs/priors/prior_fpc_1000.csv How can I get prior_fpc_1000.csv for my own dataset？ Thanks a lot in advance.

gordonjun2 commented 2 years ago

Hi there. I tested it with my own dataset already. To get the 'prior_fpc_1000.csv' for your own dataset, you need to run the train.py without it first (set your 'args.prior_path' to be None). The prior will be generated throughout the training. There are some information in the paper (eg. using 'Ignore' mode for your 'args.partial_loss_mode' for the first training).

wasdzzzzzz commented 2 years ago

Hi there. I tested it with my own dataset already. To get the 'prior_fpc_1000.csv' for your own dataset, you need to run the train.py without it first (set your 'args.prior_path' to be None). The prior will be generated throughout the training. There are some information in the paper (eg. using 'Ignore' mode for your 'args.partial_loss_mode' for the first training).

Thank you very much for your reply~