search

hgiangcao / CODs

5 stars 0 forks source link

ttm,ttt assets file missing #2

Open yolo01826 opened 10 months ago

yolo01826 commented 10 months ago

Dear author: Thank you very much for your work, but I found that some files were missing during the reproduction process. After training CODS, I wanted to train the ppo_agent for picking, but when I ran the client program you wrote, I found that the asset folder was missing the ttt format. Simulation environment file file, can you please provide the missing files?

hgiangcao commented 9 months ago

Dear @yolo01826 ,

We are transferring the data to a new server, we will soon update the link for downloading the resource.

However, our main contribution is about the consistency of object representation in different viewpoints. The demonstration of grasping/picking efficiency serves to highlight the effectiveness of utilizing object representation as input for training in robotic learning tasks. Feel free to design any robotic learning task/environment to evaluate the applicability of our object representation method.

Besides, In this work, we encountered challenges related to running V-REP in headless mode without a GUI on the server side. As a result, we've devised a solution where the client runs the V-REP environment to collect data and subsequently transmits the observations to the server for training purposes. If the PC has enough computing resources, it is not required to run client-server solution like our implementation.