experiment in RoboTHOR？

[keeper1314]

The offline data is discretized from AI2-Thor simulator. How to do the experiment in RoboTHOR environment?

[Xie-Nav]

The offline data is discretized from AI2-Thor simulator. How to do the experiment in RoboTHOR environment?

There is no recognized version of RoboTHOR's offline dataset. We collected the offline dataset on the ai2thor simulator. Due to the differences in the versions of the ai2thor simulator, we cannot guarantee that the dataset we use is exactly the same as in other works, so we decided not to make our collected dataset public for the time being.

[keeper1314]

非常感谢您百忙之中的回复！我发现很多文章在RoboTHOR环境中进行试验，但是发布的代码和数据都是关于ai2thor环境的。能否请问您RoboTHOR环境是如何进行实验的呢？我有三个猜想：（1）根据AI2环境的数据集格式，提取RoboTHOR环境的离线数据；（2）根据objectnav挑战赛的代码进行改进；https://github.com/allenai/robothor-challenge（3）掌握allenai框架，改自己的模型。https://github.com/allenai/allenact。请问哪种是采取实验的方式呢？ Thank you for your reply! I find many articles experimenting in RoboTHOR environments, but the published code and data are all about AI2 environments. Could you tell me how the experiment is carried out in the RoboTHOR environment? I have three guesses: (1) Extract the data of the RoboTHOR environment according to the data set format of the AI2 environment; (2) Make improvements according to the code of objectnav Challenge; https://github.com/allenai/robothor-challenge (3) master allenai framework, to change their own models. https://github.com/allenai/allenact

[destination-xy]

@keeper1314 @Xie-Nav 哥，你能提供个思路吗？十分感谢拉，怎么在Robothor数据集上进行训练和评估？