search

cure-lab / DeciWatch

[ECCV 2022] Official implementation of the paper "DeciWatch: A Simple Baseline for 10x Efficient 2D and 3D Pose Estimation"
Apache License 2.0
177 stars 15 forks source link

3D custom data format #6

Closed CloudedLeopard17 closed 2 years ago

CloudedLeopard17 commented 2 years ago

Hi, I am trying to run inference on 3D detection data, I have the body coordinates of people. What preprocessing do I need to do for inference? I normalised the data wrt the hip centre. I tried to visualise prepared data and the person was upside down, I am little confused. download

juxuan27 commented 2 years ago

Hi @CloudedLeopard17 , thank you for your focus. Do you mean that you want to inference your own dataset? If so, you can follow the instruction here to make your own dataset. If you only want to inference the prepared data, just follow the instruction here. For the upside-down human, we didn't deliberately place the human body like that. It is probably because the estimator's output is original upside down, and we simply copy it into our data. DeciWatch is not sensitive to the person's orientation, so if you need to generate your own dataset, you do not need to turn it to specific orientation.

CloudedLeopard17 commented 2 years ago

Hi @juxuan27 thanks for the reply, I tried to do inference on my custom data but I am getting empty predicted_pos, denoised_pos outputs in visualize.py. I am passing joints_3d inputs normalized by the hip centre. Is the joints_3d input correct?

CloudedLeopard17 commented 2 years ago

@juxuan27 I was able to run inference on my custom data. Thanks for the great work.