This document presents how to run all experiments provided in the paper and reproducing the same results.
Pre-trained models are uploaded on google drive.
To run a particular model see cmd
under log.
Name | Model Tag | Model | Log | |
---|---|---|---|---|
0 | I3D | i3d_r50_8x8 | model | log |
1 | I3D+NL | i3d_r50_nl_8x8 | model | log |
2 | Slow+NL | slow_nl_r50_8x8 | model | log |
3 | SlowFast+NL | slow_fast_nl_r50_8x8 | model | log |
4 | I3D | kpret_i3d_r50_8x8 | model | log |
5 | I3D+NL | kpret_i3d_r50_nl_8x8 | model | log |
6 | Slow+NL | kpret_slow_nl_r50_8x8 | model | log |
7 | SlowFast+NL | kpret_slow_nl_r50_8x8 | model | log |
Name | Model Tag | Model | Log | Pred | |
---|---|---|---|---|---|
0 | GPT2 | gpt2 | model | log | pred |
1 | TxD | txd | model | log | pred |
2 | SF+TxD | sfast_txd | model | log | pred |
3 | SF+TxE+TxD | sfast_txe_txd | model | log | pred |
4 | I3D+TxD | i3d_txd | model | log | pred |
5 | I3D+TxE+TxD | i3d_txe_txd | model | log | pred |
Name | Model Tag | Model | Log | |
---|---|---|---|---|
0 | Roberta Encoder | rob_evrel | model | log |
1 | Roberta Encoder | txe_evrel | model | log |
2 | Roberta with only Verbs + I3D | sfpret_vbonly_evrel | model | log |
3 | Only Video I3D | sfpret_onlyvid_evrel | model | log |
4 | Roberta + I3D | sfpret_evrel | model | log |