Overview
We offer a benchmark suite together with an evaluation server, such that authors can upload their results and get a ranking. We offer a Dataset that contains more than 50000 pictures, including 30462 images for training set, 10000 images for validation set and 10000 images for test set. If you would like to submit your results, please register, login, and follow the instructions on our submission page.
Note: We only display the highest submission of each person.
Single-Person Human Pose Estimation Track
Metrics
Followed MPII, we use PCKh evaluation measure with Person-Centric (PC) annotations.
| User Id |
Method |
|
|
PCKh |
|
Details |
Abbreviation |
Submit Time |
| 260 |
1 |
|
|
61.300 |
|
Details
| Head |
Shoulder |
Elbow |
Wrist |
Hip |
Knee |
Ankle |
UBody |
Total |
| 84.400 |
74.400 |
58.300 |
47.300 |
53.500 |
53.400 |
51.500 |
66.500 |
61.300 |
|
Abbreviation
| Contributors |
Description |
| q |
s |
|
2018-04-19 21:36:49 |
| 258 |
h |
|
|
82.500 |
|
Details
| Head |
Shoulder |
Elbow |
Wrist |
Hip |
Knee |
Ankle |
UBody |
Total |
| 93.100 |
89.600 |
84.500 |
82.400 |
69.800 |
77.300 |
77.700 |
87.500 |
82.500 |
|
Abbreviation
|
2018-05-18 04:23:34 |
| 288 |
JDAI-human |
|
|
90.900 |
|
Details
| Head |
Shoulder |
Elbow |
Wrist |
Hip |
Knee |
Ankle |
UBody |
Total |
| 95.900 |
94.800 |
92.300 |
90.400 |
81.400 |
90.300 |
90.200 |
93.400 |
90.900 |
|
Abbreviation
| Contributors |
Description |
| Fangyu Li,
Wu Liu, Qian Bao, Peng Cheng, Tao Mei. JD AI Research. |
Our final method is based on cascading pyramid network [1], stacked hourglass [2] and RMPE [3]. We modified and fine-tuned the above models on the LIP dataset with increasing the size of inputs and feature maps, data augmentation, hard point mining, and other strategies. Finally, we fused the models to get the best results.
[1] Chen, Yilun, et al. "Cascaded Pyramid Network for Multi-Person Pose Estimation." arXiv preprint arXiv:1711.07319 (2017).
[2] Newell, Alejandro, et al. "Stacked hourglass networks for human pose estimation." European Conference on Computer Vision. 2016.
[3] Fang, Hao-Shu, et al. "RMPE: Regional multi-person pose estimation." The IEEE International Conference on Computer Vision. 2017. |
|
2018-06-10 03:27:17 |
| 283 |
|
|
|
83.300 |
|
Details
| Head |
Shoulder |
Elbow |
Wrist |
Hip |
Knee |
Ankle |
UBody |
Total |
| 93.600 |
89.800 |
85.000 |
83.700 |
70.400 |
78.800 |
78.900 |
88.100 |
83.300 |
|
Abbreviation
|
2018-05-24 09:21:20 |
| 298 |
ByteDance-SEU-Baseline |
|
|
90.200 |
|
Details
| Head |
Shoulder |
Elbow |
Wrist |
Hip |
Knee |
Ankle |
UBody |
Total |
| 95.800 |
94.400 |
91.700 |
89.600 |
80.200 |
89.500 |
89.200 |
93.000 |
90.200 |
|
Abbreviation
| Contributors |
Description |
| Kai Su(Southeast University & ByteDance AI Lab), Dongdong Yu(ByteDance AI Lab), Zhenqi Xu(ByteDance AI Lab), Xin Geng(Southeast University), Changhu Wang(ByteDance AI Lab) |
We revised and finetuned the Cascaded Pyramid Network on the LIP dataset, then ensembled two different backbone models (resnet-50 & resnet-101). [1] Chen, Yilun, et al. "Cascaded Pyramid Network for Multi-Person Pose Estimation." arXiv preprint arXiv:1711.07319 (2017). |
|
2018-06-08 02:54:49 |
