# Action Recognition on NTU RGB+D 120 (XSub & XSet)

Source: [Action Recognition on NTU RGB+D 120 (XSub & XSet)](<https://www.sota2.com/research/sota/action-recognition-on-ntu-rgb-plus-d-120-xsub-and-xset>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-26T06:07:16.262Z

Evaluation of video action recognition performance using the NTU RGB+D 120 dataset, reporting results across both the Cross-Subject (XSub) and Cross-Set (XSet) evaluation protocols.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Action Recognition |
| Dataset (including any stated split) | NTU RGB+D 120 |
| Metric | Top-1 Accuracy (XSub) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 11 of 11 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Top-1 Accuracy (XSub)](<https://www.sota2.com/research/sota/action-recognition-on-ntu-rgb-plus-d-120-xsub-and-xset.md?metric=Top-1%20Accuracy%20(XSub)>): Higher is better; unit: PERCENT.

- [Top-1 Accuracy (XSet)](<https://www.sota2.com/research/sota/action-recognition-on-ntu-rgb-plus-d-120-xsub-and-xset.md?metric=Top-1%20Accuracy%20(XSet)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Top-1 Accuracy (XSub) | Result date | Report paper |
| --- | --- | --- | --- |
| MMNet (Inception-v3) | 92.9 | Not recorded | 68fa77871e209131cea73d90 |
| EPAM-Net (Our proposed approach) | 92.4 | Not recorded | 68fa77871e209131cea73d90 |
| π-ViT | 91.9 | Not recorded | 68fa77871e209131cea73d90 |
| VPN++ + 3D Poses | 90.7 | Not recorded | 68fa77871e209131cea73d90 |
| TSMF | 87 | Not recorded | 68fa77871e209131cea73d90 |
| MS-G3D | 86.9 | Not recorded | 68fa77871e209131cea73d90 |
| VPN++ | 86.7 | Not recorded | 68fa77871e209131cea73d90 |
| VPN (I3D) | 86.3 | Not recorded | 68fa77871e209131cea73d90 |
| PoseConv3D | 86 | Not recorded | 68fa77871e209131cea73d90 |
| 2s-AGCN | 82.9 | Not recorded | 68fa77871e209131cea73d90 |
| ST-GCN | 79 | Not recorded | 68fa77871e209131cea73d90 |

### MMNet (Inception-v3)

Top-1 Accuracy (XSub): 92.9. Result ID: 69953dde3badfaffe626d9c7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77871e209131cea73d90.md>)

Recorded evaluation configuration: {"Modality":"Multimodal (Skeleton+RGB)","Backbone":"Inception-v3","GFLOPs":"89.2","Param (M)":"34.2"}. This record does not assert independent reproduction.

### EPAM-Net (Our proposed approach)

Top-1 Accuracy (XSub): 92.4. Result ID: 69953dde3badfaffe626d9c9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77871e209131cea73d90.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77871e209131cea73d90.md>)

Recorded evaluation configuration: {"Modality":"Multimodal (Skeleton+RGB)","GFLOPs":"8.1","Param (M)":"2.5"}. This record does not assert independent reproduction.

### π-ViT

Top-1 Accuracy (XSub): 91.9. Result ID: 69953dde3badfaffe626d9c8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77871e209131cea73d90.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa715c1e209131cea50dad.md>)

[Implementation](<https://github.com/dominickrei/pi-vit/>)

Recorded evaluation configuration: {"Modality":"RGB","Skeleton usage":"Training only","Input clips per video":"10","GFLOPs":"590.0","Param (M)":"121.4"}. This record does not assert independent reproduction.

### VPN++ + 3D Poses

Top-1 Accuracy (XSub): 90.7. Result ID: 69953dde3badfaffe626d9c6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77871e209131cea73d90.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69ef1e209131cea25059.md>)

[Implementation](<https://github.com/srijandas07/vpnplusplus>)

Recorded evaluation configuration: {"Modality":"Multimodal (Skeleton+RGB)","GFLOPs":"125.8","Param (M)":"15.5"}. This record does not assert independent reproduction.

### TSMF

Top-1 Accuracy (XSub): 87. Result ID: 69953dde3badfaffe626d9c3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77871e209131cea73d90.md>)

Recorded evaluation configuration: {"Modality":"Multimodal (Skeleton+RGB)","GFLOPs":"85.4","Param (M)":"20.8"}. This record does not assert independent reproduction.

### MS-G3D

Top-1 Accuracy (XSub): 86.9. Result ID: 69953dde3badfaffe626d9c1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77871e209131cea73d90.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa678c1e209131cea16e11.md>)

[Implementation](<https://www.sota2.com/github.com/kenziyuliu/ms-g3d>)

Recorded evaluation configuration: {"Modality":"Skeleton","GFLOPs":"16.7","Param (M)":"2.8"}. This record does not assert independent reproduction.

### VPN++

Top-1 Accuracy (XSub): 86.7. Result ID: 69953dde3badfaffe626d9c5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77871e209131cea73d90.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa69ef1e209131cea25059.md>)

[Implementation](<https://github.com/srijandas07/vpnplusplus>)

Recorded evaluation configuration: {"Modality":"RGB","Skeleton usage":"Training only"}. This record does not assert independent reproduction.

### VPN (I3D)

Top-1 Accuracy (XSub): 86.3. Result ID: 69953dde3badfaffe626d9c4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77871e209131cea73d90.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa67ec1e209131cea1928d.md>)

[Implementation](<https://github.com/srijandas07/VPN>)

Recorded evaluation configuration: {"Modality":"Multimodal (Skeleton+RGB)","Backbone":"I3D","GFLOPs":"107.9","Param (M)":"24.0"}. This record does not assert independent reproduction.

### PoseConv3D

Top-1 Accuracy (XSub): 86. Result ID: 69953dde3badfaffe626d9c2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77871e209131cea73d90.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6c321e209131cea32c30.md>)

[Implementation](<https://github.com/kennymckormick/pyskl>)

Recorded evaluation configuration: {"Modality":"Skeleton","GFLOPs":"15.9","Param (M)":"2.0"}. This record does not assert independent reproduction.

### 2s-AGCN

Top-1 Accuracy (XSub): 82.9. Result ID: 69953dde3badfaffe626d9c0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77871e209131cea73d90.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa65fa1e209131cea0d9f3.md>)

[Implementation](<https://github.com/lshiwjx/2s-AGCN>)

Recorded evaluation configuration: {"Modality":"Skeleton","GFLOPs":"8.8","Param (M)":"3.5"}. This record does not assert independent reproduction.

### ST-GCN

Top-1 Accuracy (XSub): 79. Result ID: 69953dde3badfaffe626d9bf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77871e209131cea73d90.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa63f21e209131cea01b83.md>)

[Implementation](<https://github.com/yysijie/st-gcn>)

Recorded evaluation configuration: {"Modality":"Skeleton","GFLOPs":"3.8","Param (M)":"3.1"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
