mxnet_Realtime_Multi-Person_Pose_Estimation
mxnet_Realtime_Multi-Person_Pose_Estimation copied to clipboard
How to understand for the "center_map" in the prototxt? What's the meaning for the networks?
layer { name: "image" type: "Slice" bottom: "data" top: "image" top: "center_map" slice_param { slice_point: 3 axis: 1 } } layer { name: "silence2" type: "Silence" bottom: "center_map" } @dragonfly90 Thanks? Looking forward your attention
@Ai-is-light I think it means to crop the image based on the center of the person, but I am not sure.
I have seen that some also did think the mask and ground truth are based on the person and their keypoints, maybe you are right. Thanks
how do get the first 2644 images in the val2014 to test the model you trianed ? just use the code of author's genJOSN? I mean, whether is other method to get the more typical small val-dataset?@d @dragonfly90