metadata
license: apache-2.0
tags:
- video-classification
- timesformer
library_name: transformers
TimeSformer fine-tuned model
Checkpoint: timesformer_max_full_even_6.pt
Training configuration
- Model: TimeSformer-Base (Kinetics-400 pretrained)
- Frame sampling: even
- Data balancing: max_full
- Unfrozen last layers: 6
- Task: workout action classification (plank, push-up, squat, russian-twist)
Notes
This repository provides a raw PyTorch checkpoint.
To use it, load the checkpoint into a TimesformerForVideoClassification
model with the same configuration used during training.