search

geopavlakos / ordinal-pose3d

Code for "Ordinal Depth Supervision for 3D Human Pose Estimation", CVPR 2018
111 stars 16 forks source link

lsp_mpii_ordinal data format #4

Open Fangyh09 opened 6 years ago

Fangyh09 commented 6 years ago

 I downloaded the data provided from http://visiondata.cis.upenn.edu/ordinal/lsp-mpii/lsp-mpii-ordinal.zip. However, I found the joints annotation does not match the image. Does joints ordinates are scaled to 256x256? Thanks!

data_dir = "lsp-mpii-ordinal/mpii_upis1h"
img_path = os.path.join(data_dir, "images/12172_full.jpg")
joints_mat = matio.loadmat(os.path.join(data_dir,
                                             "joints.mat"))['joints']
ordinal_mat = matio.loadmat(os.path.join(data_dir,
                                              "ordinal.mat"))['ord']        
img = io.imread(img_path)
height, width, _ = img.shape
idx = 12171
joints = joints_mat[:, :, idx]
lvandoit commented 4 years ago

hi @Fangyh09 ,Will the network predict the volume heatmap and then output the marginal heatmap？

zhLawliet commented 2 years ago

hi @Fangyh09 can you share the datasets http://visiondata.cis.upenn.edu/ordinal/lsp-mpii/lsp-mpii-ordinal.zip ?