Skip to content

Latest commit

 

History

History
35 lines (24 loc) · 6.63 KB

File metadata and controls

35 lines (24 loc) · 6.63 KB

Pretrained Models

Introduction

We present the detailed performance on various datasets and modalities.

All the checkpoints and training logs are provided in the Google Drive. We sincerely hope that this repo could be helpful for your research.

Experimental Results

The detailed results for pretrained models are displayed below:

Modality NTU 60 X-Sub NTU 60 X-View NTU 120 X-Sub NTU 120 X-Set
Joint 92.75 97.39 87.39 89.59
Bone 93.01 96.94 90.01 91.57
K-Bone 93.01 97.01 89.49 90.46
2-ensemble 94.01 97.73 90.51 92.43
4-ensemble 94.30 97.99 91.35 92.97
6-ensemble 94.51 98.19 91.70 93.05
Modality Kinetics-Skeleton FineGYM UAV-Human CSv1 UAV-Human CSv2
Joint 50.69 94.64 47.25 73.66
Bone 49.10 95.70 48.50 73.98
K-Bone 48.26 95.54 47.44 73.42
2-ensemble 52.24 96.08 50.31 76.00
4-ensemble 52.98 96.46 50.85 76.95
6-ensemble 53.57 96.53 52.05 78.07

We adopt the widely-used six-stream ensemble strategy introduced in InfoGCN. Here K-Bone denotes the newly skeleton representation proposed by InfoGCN. Interestingly, we find that the improvement of multi-stream ensemble method mainly comes from complementarity and stochasticity. For well-performing models, stochastic boosting of single-modality is more efficient than complementary boosting of motion-modality. The detailed comparisons for various datasets are provided in {dataset}_ensemble.py.

In addition, we use three augmentation techniques Flip, Part Drop, and Mixup that could provide performance gains. Here, Mixup doubles the samples, which leads to longer training times. Nevertheless, when coupled with multi-modal semantic priors, it could deliver superior performance. Notably, due to randomness, these augmentations may also make the performance fluctuate. You could choose whether or not to use them based on the actual needs.