Phase Classification on Penn Action (test)
86.05AccuracySupervised Learning
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Supervised LearningBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=1.02019.04 | 86.05 | — | — | — | |
| TCC + TCNBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=1.02019.04 | 85.22 | — | — | — | |
| TCC + TCNBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=0.52019.04 | 84.79 | — | — | — | |
| LAV2021.03 | 84.47 | — | — | — | |
| TCCBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=1.02019.04 | 84.45 | — | — | — | |
| TCC + SaLBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=1.02019.04 | 84.29 | — | — | — | |
| TCC + TCNBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=0.12019.04 | 84.27 | — | — | — | |
| LAV + TCC2021.03 | 84.14 | — | — | — | |
| TCNBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=1.02019.04 | 84.04 | — | — | — | |
| TCC2021.03 | 83.94 | — | — | — | |
| TCNBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=0.52019.04 | 83.67 | — | — | — | |
| TCC + SaLBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=0.52019.04 | 83.46 | — | — | — | |
| TCCBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=0.52019.04 | 83.35 | — | — | — | |
| TCN2021.03 | 82.99 | — | — | — | |
| Supervised LearningBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=0.52019.04 | 82.78 | — | — | — | |
| SAL2021.03 | 82.03 | — | — | — | |
| TCNBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=0.12019.04 | 81.99 | — | — | — | |
| TCC + SaLBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=0.12019.04 | 81.93 | — | — | — | |
| TCCBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=0.12019.04 | 81.26 | — | — | — | |
| Supervised Learning% of Labels=1.0, Backbone=VGG-M, Training Protocol=from scratch2019.04 | 79.98 | — | — | — | |
| SaLBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=1.02019.04 | 79.96 | — | — | — | |
| LAVJoint all-action model=true, Shared single encoder=true, Classifier=13 SVMs2021.03 | 78.68 | — | — | — | |
| SaLBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=0.52019.04 | 78.26 | — | — | — | |
| TCC% of Labels=1.0, Backbone=VGG-M, Training Protocol=from scratch2019.04 | 77.3 | — | — | — | |
| TCC% of Labels=0.5, Backbone=VGG-M, Training Protocol=from scratch2019.04 | 76.39 | — | — | — | |
| SaLBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=0.12019.04 | 74.87 | — | — | — | |
| TCC% of Labels=0.1, Backbone=VGG-M, Training Protocol=from scratch2019.04 | 74.68 | — | — | — | |
| TCCJoint all-action model=true, Shared single encoder=true, Classifier=13 SVMs2021.03 | 74.39 | — | — | — | |
| Supervised Learning% of Labels=0.5, Backbone=VGG-M, Training Protocol=from scratch2019.04 | 72.86 | — | — | — | |
| SaL% of Labels=1.0, Backbone=VGG-M, Training Protocol=from scratch2019.04 | 72.53 | — | — | — | |
| TCN% of Labels=1.0, Backbone=VGG-M, Training Protocol=from scratch2019.04 | 72.15 | — | — | — | |
| TCN% of Labels=0.5, Backbone=VGG-M, Training Protocol=from scratch2019.04 | 71.41 | — | — | — | |
| SaL% of Labels=0.5, Backbone=VGG-M, Training Protocol=from scratch2019.04 | 71.1 | — | — | — | |
| TCN% of Labels=0.1, Backbone=VGG-M, Training Protocol=from scratch2019.04 | 69.65 | — | — | — | |
| SALJoint all-action model=true, Shared single encoder=true, Classifier=13 SVMs2021.03 | 68.15 | — | — | — | |
| TCNJoint all-action model=true, Shared single encoder=true, Classifier=13 SVMs2021.03 | 68.09 | — | — | — | |
| Supervised LearningBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=0.12019.04 | 67.1 | — | — | — | |
| SaL% of Labels=0.1, Backbone=VGG-M, Training Protocol=from scratch2019.04 | 66.15 | — | — | — | |
| ImageNet FeaturesBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=1.02019.04 | 52.86 | — | — | — | |
| ImageNet FeaturesBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=0.52019.04 | 50.91 | — | — | — | |
| Supervised Learning% of Labels=0.1, Backbone=VGG-M, Training Protocol=from scratch2019.04 | 50.71 | — | — | — | |
| Random FeaturesBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=1.02019.04 | 46.81 | — | — | — | |
| Random FeaturesBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=0.52019.04 | 46.19 | — | — | — | |
| ImageNet FeaturesBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=0.12019.04 | 44.96 | — | — | — | |
| Random FeaturesBackbone=ResNet-50, Pre-trained=ImageNet, Evaluation Protocol=Fine-tuning, % of Labels=0.12019.04 | 44.18 | — | — | — | |
| CASAPose=true, ImageNet pre-trained=false2022.04 | — | 88.55 | 91.87 | 92.2 | |
| LAVPose=false, ImageNet pre-trained=true2022.04 | — | 83.56 | 83.95 | 84.25 | |
| LAVPose=true, ImageNet pre-trained=false2022.04 | — | 79.83 | 80.2 | 80.2 | |
| SaLPose=false, ImageNet pre-trained=true2022.04 | — | 74.87 | 78.26 | 79.96 | |
| TCCPose=false, ImageNet pre-trained=true2022.04 | — | 79.72 | 81.11 | 81.35 | |
| TCCPose=true, ImageNet pre-trained=false2022.04 | — | 79.53 | 83.75 | 84.51 | |
| TCNPose=false, ImageNet pre-trained=true2022.04 | — | 81.99 | 83.67 | 84.04 |