COMEDIAN: Self-Supervised Learning and Knowledge Distillation for Action Spotting using Transformers
Paper
•
2309.01270
•
Published
COMEDIAN ViSwin Tiny checkpoints
Context
These 3 models were obtained following the COMEDIAN method to perform the action spotting on the SoccerNet-v2 dataset using the ViSwin Tiny architecture.
They were trained on three different seeds (42, 203, 666) reflected by the checkpoints' filename.
How to use
To use these checkpoints please refer to the eztorch library and in particular the COMEDIAN documentation.
Inference Providers
NEW
This model isn't deployed by any Inference Provider.
🙋
Ask for provider support