yifan-you-37 / omnihang Goto Github PK

Dear Yifan,

I am very interested in this project and tried to replicate your code by running the file 's3_rl_collect.py' and got a .json file as output. I am not sure about the meaning of each elements in the output. I want to get the final relative pose between hanger and supporting item and success rate. I think that "succ" means whether the hanging of this pair objects is successful, "final_pred_transl" means the translation between two objects and "final_pred_aa" is the orientation between two objects (as marked in the image below)?

Thank you very much!

Hi
I'm very interested in your work and would like to reproduce your results.
I downloaded all the data and code, but due to the lack in instructions and documentation, I find it difficult to figure out which scripts should be run in which order and with which arguments. I'd very much appreciate any additional documentation on
-how to visualize the final result, given the pretrained model (similar to the gifs on the project webpage)
-how to train a new model from scratch given the data
-how to generate additional training data

Thank you very much for your consideration.