Closed Alidaling closed 1 year ago
Yes, the model is generally named with [model variant]-[dataset]
. The default setting is given at the document: https://iqa-pytorch.readthedocs.io/en/latest/benchmark.html#performance-evaluation-protocol.
For your question, topiq_fr
is trained with kadid dataset, and topiq_fr-pipal
is trained with pipal dataset.
Are the training data used for the topiq_fr model and the topiq_fr-pipal model different? What are the specific datasets used to train them?Thanks!