anosorae / IRRA

Cross-Modal Implicit Relation Reasoning and Aligning for Text-to-Image Person Retrieval (CVPR 2023)
MIT License
183 stars 25 forks source link

The accuracy of IDLoss experiment is much lower than 65.33% of the paper #16

Closed xie-xue closed 1 year ago

xie-xue commented 1 year ago

parameter loss_names='id' lr=1e-05 lr_factor=5.0 Concrete result image Looking forward to your reply!

anosorae commented 1 year ago

The id loss in the table is not a separate loss, but is trained jointly with the itc loss in the baseline.

xie-xue commented 1 year ago

Thank you