Do we really need temporal convolutions in action segmentation?

Benchmark Model Rank Results
action-segmentation-on-50-salads-1EUT#6F1@50%: 81F1@25%: 87.5F1@10%: 89.2Acc: 87.4Edit: 82.9
action-segmentation-on-breakfast-1EUT#7Average F1: 69.3F1@50%: 59.8F1@25%: 71.8F1@10%: 76.2Edit: 74.6
action-segmentation-on-gtea-1EUT#18F1@50%: 74F1@25%: 87.2F1@10%: 88.2Acc: 77Edit: 83.9