Pre-trained models

Pre-Trained model nomination list

There are many approaches that teams could take in the SPEAR Challenge and, of course, machine learning is a very promising one. It is well known that systems generally perform better when trained on more data. The Datasets provided in the SPEAR Challenge are already big but they are still limited. To allow for re-use/adaption of previously developed methods, the SPEAR Challenge allowed teams to nominate up to 5 pre-trained models. In order to keep everything fair and allow other teams to also benefit from these pre-trained models, we required that the pre-trained models be publicly available, that is,

• the model weights are available to download, and
• the data that it was trained on have a permissive licence.

We are now pleased to announce that the following models have been approved for use in the SPEAR Challenge.