search

mks0601 / 3DMPPE_POSENET_RELEASE

Official PyTorch implementation of "Camera Distance-aware Top-down Approach for 3D Multi-person Pose Estimation from a Single RGB Image", ICCV 2019
MIT License
807 stars 147 forks source link

Processing of raw datasets #101

Open sainan-zhang opened 2 years ago

sainan-zhang commented 2 years ago

It is a very excellent work! Thanks for your detailed sharing. I have learned your codes of ROOTNET and POSENET and I want to make some improvements based on your networks. But I don't know how you process the original dataset into the form you shared. In your datasets, such as Human36M, the image files are stored in many separate folders. Can I treat the image in each folder as each frame of an action video? Looking forward to your reply! Thanks!

mks0601 commented 2 years ago

Yes you can treat the image in each folder as each frame of an action video. All information about data are included in the provided json files.

sainan-zhang commented 2 years ago

Thanks very much!

YHDang commented 2 years ago

It is a very excellent work! Thanks for your detailed sharing. I have learned your codes of ROOTNET and POSENET and I want to make some improvements based on your networks. But I don't know how you process the original dataset into the form you shared. In your datasets, such as Human36M, the image files are stored in many separate folders. Can I treat the image in each folder as each frame of an action video? Looking forward to your reply! Thanks!

Hello, I have some trouble downloading the shared dataset. So I want to process the original H36M dataset. But the original dataset only contains videos. If I extract frames from the video, what should the frame rate be, 30fps?