PeterGuoRuc/RucEmo-muse

Workspace For Muse2020 challenge

run this project

requirementes

• python==3.7

• pytorch>=1.0.0

using scripts to run experiments

• run sh scripts/train_fcfusion arousal/valence 1 and modify --feature_set args

result

basic result on validation set

Features		arousal	valence
Uni-modal	lld	0.3841	0.03
	wav2vec	0.3092	0.1737
	denseface	0.2954	0.0524
	vggface	0.2996	0.0852
	bert_base	0.1643	0.3131
multi-modal	lld-bert-glove	0.3378	0.3447
	wav2vec-bert_base-glove	0.3440	0.3556
	lld-wav2vec-denseface-au-bert	0.4670	0.3571
	lld-wav2vec-vgg16-au-bert	0.4514	0.3153

Note that is table is not complete yet.

submission result

name	arousal	valence
baseline	0.2834	0.2431
submit1	0.3082	40.814
submit2	0.3753	0.4194
submit3	0.4346	0.4513