Action Recognition on SBU Kinect Interaction Dataset (5-fold cross val)
98.3AccuracyVA-fusion
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| VA-fusionaugmentation=true2018.04 | 98.3 | — | — | — | — | — | |
| VA-RNNaugmentation=true2018.04 | 97.5 | — | — | — | — | — | |
| VA-CNNaugmentation=true2018.04 | 95.7 | — | — | — | — | — | |
| GCA-LSTM2018.04 | 94.1 | — | — | — | — | — | |
| Clips+CNN+MTLN2018.04 | 93.6 | — | — | — | — | — | |
| Clips + CNN + MTLNinput=Clips, aggregation=MTLN2017.03 | 93.57 | — | — | — | — | — | |
| ST-LSTM + Trust Gatemechanism=Trust Gate2017.03 | 93.3 | — | — | — | — | — | |
| ST-LSTM + Trust Gate2018.04 | 93.3 | — | — | — | — | — | |
| Clips + CNN + Concatenationinput=Clips, aggregation=Concatenation2017.03 | 92.86 | — | — | — | — | — | |
| Clips + CNN + Poolinginput=Clips, aggregation=Pooling2017.03 | 92.26 | — | — | — | — | — | |
| STA-LSTM2018.04 | 91.5 | — | — | — | — | — | |
| Frames + CNNinput=Frames, method=CNN2017.03 | 90.88 | — | — | — | — | — | |
| Deep LSTM + Co-occurrencemechanism=Co-occurrence2017.03 | 90.41 | — | — | — | — | — | |
| Co-occurrence RNN2018.04 | 90.4 | — | — | — | — | — | |
| ST-LSTM2017.03 | 88.6 | — | — | — | — | — | |
| Joint Feature2017.03 | 86.9 | — | — | — | — | — | |
| Joint featuresource_paper=[16]2018.04 | 86.9 | — | — | — | — | — | |
| Deep LSTM2017.03 | 86.03 | — | — | — | — | — | |
| CHARM2017.03 | 83.9 | — | — | — | — | — | |
| HBRNN-L2018.04 | 80.4 | — | — | — | — | — | |
| Hierarchical RNN2017.03 | 80.35 | — | — | — | — | — | |
| Joint featuresource_paper=[56]2018.04 | 80.3 | — | — | — | — | — | |
| Coordinates + FTPinput=Coordinates, method=FTP2017.03 | 79.75 | — | — | — | — | — | |
| Raw skeletonsource_paper=[16]2018.04 | 79.4 | — | — | — | — | — | |
| Raw Skeleton2017.03 | 49.7 | — | — | — | — | — | |
| Raw skeletonsource_paper=[56]2018.04 | 49.7 | — | — | — | — | — | |
| LSTMSupervision Type=Supervised, Evaluation Protocol=5-fold Cross Validation2020.08 | 47.1 | 49.1 | 53.2 | 37.5 | 42 | 53.8 | |
| CAE+Supervision Type=Unsupervised, Evaluation Protocol=5-fold Cross Validation, Approach=AS-CAL2020.08 | 46.7 | 52.7 | 50 | 44.6 | 37 | 49.2 | |
| CAESupervision Type=Unsupervised, Evaluation Protocol=5-fold Cross Validation, Approach=AS-CAL2020.08 | 42.6 | 52.7 | 46.2 | 41.1 | 31.5 | 41.5 | |
| GRUSupervision Type=Supervised, Evaluation Protocol=5-fold Cross Validation2020.08 | 36.5 | 40 | 40.4 | 28.6 | 33.3 | 40 | |
| RNNSupervision Type=Supervised, Evaluation Protocol=5-fold Cross Validation2020.08 | 34.5 | 40 | 42.3 | 26.8 | 27.8 | 35.4 |