Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

PoseFormerV2: Exploring Frequency Domain for Efficient and Robust 3D Human Pose Estimation

About

Recently, transformer-based methods have gained significant success in sequential 2D-to-3D lifting human pose estimation. As a pioneering work, PoseFormer captures spatial relations of human joints in each video frame and human dynamics across frames with cascaded transformer layers and has achieved impressive performance. However, in real scenarios, the performance of PoseFormer and its follow-ups is limited by two factors: (a) The length of the input joint sequence; (b) The quality of 2D joint detection. Existing methods typically apply self-attention to all frames of the input sequence, causing a huge computational burden when the frame number is increased to obtain advanced estimation accuracy, and they are not robust to noise naturally brought by the limited capability of 2D joint detectors. In this paper, we propose PoseFormerV2, which exploits a compact representation of lengthy skeleton sequences in the frequency domain to efficiently scale up the receptive field and boost robustness to noisy 2D joint detection. With minimum modifications to PoseFormer, the proposed method effectively fuses features both in the time domain and frequency domain, enjoying a better speed-accuracy trade-off than its precursor. Extensive experiments on two benchmark datasets (i.e., Human3.6M and MPI-INF-3DHP) demonstrate that the proposed approach significantly outperforms the original PoseFormer and other transformer-based variants. Code is released at \url{https://github.com/QitaoZhao/PoseFormerV2}.

Qitao Zhao, Ce Zheng, Mengyuan Liu, Pichao Wang, Chen Chen• 2023

Related benchmarks

TaskDatasetResultRank
3D Human Pose EstimationMPI-INF-3DHP (test)
PCK97.9
559
3D Human Pose EstimationHuman3.6M (test)
MPJPE (Average)35.6
547
3D Human Pose EstimationHuman3.6M (Protocol #1)
MPJPE (Avg.)45.2
440
3D Human Pose EstimationHuman3.6M (Protocol 2)
Average MPJPE35.64
315
3D Human Pose EstimationHuman3.6M Protocol 1 (test)
Dir. Error (Protocol 1)41.26
183
3D Human Pose EstimationHuman3.6M (subjects 9 and 11)
Average Error45.2
180
3D Human Pose EstimationHuman3.6M
MPJPE45.2
160
3D Human Pose Estimation3DPW
PA-MPJPE56.2
119
3D Human Pose EstimationMPI-INF-3DHP
PCK97.9
108
3D Human Pose EstimationHuman3.6M (S9, S11)
Average Error (MPJPE Avg)35.5
94
Showing 10 of 13 rows

Other info

Code

Follow for update